2814 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 Understanding DAOs: An Empirical Study on Governance Dynamics Qin Wang , Member, IEEE, Guangsheng Yu , Member, IEEE, Yilin Sai , Student Member, IEEE, Caijun Sun , Lam Duc Nguyen , Member, IEEE, and Shiping Chen , Senior Member, IEEE Abstract—As a typical instance of human–computer interac- TABLEI tion,thenotionofdecentralizedautonomousorganization(DAO) THISWORKVERSUSDAOSTUDIES representsanorganizationconstructedbyautomaticallyexecuted rules, such as via smart contracts, incorporating features of the Examples Method Target In-time Applicable Scale permissionless committee, transparent proposals, and fair con- [2][3][4][5] Literaturereview Publications Notvery n/a <30 tributions by stakeholders. As of May 2023, DAO has impacted [6][7] Framework Properties n/a Priori n/a over $24.3B market caps. However, there are limited studies focused on this emerging field. To fill the gap, we start from the [8][9][10] Empiricalstudy Projects In-time Practical <22 ground truth by empirically studying the breadth and depth of Thiswork Empiricalstudy Launchpad In-time Practical >500 the DAO markets in mainstream public chain ecosystems in this article.WediveintothemostwidelyadoptableDAOlaunchpad, Snapshot, which covers 95% of the wild DAO projects for data andactivitiestobetransparentandtraceable.Stakeholdersare collectionandanalysis.ByintegratingextensivelyenrolledDAOs eligibletopropose,vote,andenactchangesforDAOproposals. and corresponding data measurements, we explore statistical re- sourcesfromSnapshotandanalyzedatafrom581DAOprojects, AsofMay2023,atotalof12824organizationshavebeencre- encompassing 16246 proposals over the course of 3+ years. ated.TheinvestedfundingtowardtheseDAOs(a.k.a.,treasury) Our empirical research has uncovered a multitude of previously has reached up to $24.3B, while engaged members increased unknownfactsaboutDAOs,spanningtopicssuchastheirstatus, 531x from 13K to 6.9M members during the last 6 years1. features, performance, threats, and ways of improvement. We Among them, 4.5M participants are active voters or proposal have distilled these findings into a series of key insights and takeaway messages, emphasizing their significance. Notably, our makers.DAOsaccordinglybecomeaforcetobereckonedwith study is the first of its kind to comprehensively examine the intheWeb3 space[1]andnew cryptocurrency markets. DAOecosystemwithafocusonscaleandscopeofdata,real-time The Open Problem: Although the concept of DAOs has relevance,practicalimplementations,andcomprehensivemetrics, gained traction in recent years, their structural development is addressing critical gaps in the current literature. still in its nascent stage. One of the primary challenges is that Index Terms—Blockchain, decentralized autonomous organi- DAOprojectsarediverse,withvaryingobjectivesandfunction- zation (DAO), human–computer interaction (HCI), snapshot. alities.SomeDAOsbeginwithaclearpurpose,suchasUniswap and Bancor, which can remain focused on serving a specific I. INTRODUCTION communityorusers.Incontrast,otherDAOsmaydivergefrom DECENTRALIZED autonomous organizations (DAOs) their initial objectives. This means a DAO may begin with a simple goal such as collecting NFTs, and then morph into a emergewiththerapiddevelopmentofcryptocurrencyand community to attract participants (e.g., PleasrDAO), a trading blockchain.DAOisanentitythatiscollaborativelymanagedby platform to trade NFTs (Opensea), or an incubator to invest on-chainparticipantstodeployresources,releaseproposalsand artists (BAYC). The variety of forms and outcomes can be make decisions. The usage of DAO in governance can decen- confusing for newcomers and experts alike in the blockchain tralizetheoperationviablockchainbyenablingon-chainrules space.Creatingacomprehensive andstructuralviewofDAOs isstillasignificantchallengethatrequiresfurtherdevelopment and refinement. Recent studies attempt to examine DAOs (cf. Table I). Sev- Received 23 October 2023; revised 21 November 2024 and 21 January eral studies (see Row 1) begin by providing an overview of 2025;accepted5February2025.Dateofpublication17February2025;date existing literature. However, due to the delay in publication, ofcurrentversion2October2025.(Correspondingauthor:QinWang.) QinWang,YilinSai,LamDucNguyen,andShipingChenarewithCSIRO academic works may lack current and persuasive examples of Data61, Sydney, NSW 2015, Australia (e-mail: Qin.Wang@data61.csiro.au; DAOprojects,leadingtooutdatedinformation(equiv.in-timeor Yilin.Sai@data61.csiro.au; Lam.Nguyen@data61.csiro.au; Shiping.Chen@ not).Otherworks(seeRow2)proposeahigh-levelframework data61.csiro.au). Guangsheng Yu is with The University of Technology Sydney, Sydney, to discuss DAO properties, but their abstracted metrics are NSW2007,Australia(e-mail:Guangsheng.Yu@uts.edu.au). created before events, making them impractical (close-to-real Caijun Sun is with Zheliang Lab, Hangzhou, Zhejiang Province 311121, P.R.China(e-mail:sun.cj@zhejianglab.com). DigitalObjectIdentifier10.1109/TCSS.2025.3539889 1Datasource:https://deepdao.io/organizations.[May2023] ©2025TheAuthors.ThisworkislicensedunderaCreativeCommonsAttribution4.0License.Formoreinformation, seehttps://creativecommons.org/licenses/by/4.0/ WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2815 or applicable?). Several studies (Row 3) focus on capturing featuresfromrealDAOprojectsthroughempiricalresearch,but theirsamplepoolsareoftenlimited(resourcefulorinvestigated projectsinlargescale?).Alloftheseeffortsseemtofallshort ofprovidingacomprehensiveandup-to-dateunderstandingof DAOsforreaders (moreinTable III). Our Attempts:Toaddresstheaforementionedshortcomings, wehavedevisedauniqueapproachtoourstudy.Afterconduct- ingthoroughresearch,wehavefoundthatexistingDAOlaunch- pads and DataFeeds have amassed a wealth of information on numerous DAO projects, including both long-standing DAOs thathaveceasedoperationsandnewly-launchedonesthathave appearedwithinthelastmonth.Inanefforttoavoidduplicating Fig.1. OverviewofthisWork. theworkofothers,wehaveoptedtoomitDataFeedsandsome launchpadsthathavealreadypresentedanalyzeddata.Instead, we are focusing our attention on a lesser-known launchpad, TABLEII namelySnapshot[11],whichhascompiledasignificantnumber RESULTGUIDANCE ofDAOs buthas yettoconduct extensive analyses on them. Index Description The emergence of Snapshot is to overcome the issue of highcostsassociatedwithon-chainoperationsduetothecom- plexity of consensus and frequent voter interactions. Snapshot accordingly introduced an off-chain voting tool that enables practitioners to efficiently access popular DAOs for voting, managing, auditing, and researching. Snapshot serves as a launchpadthatcapturesover95%ofin-the-wildDAOprojects (over11000spaces)andoffersopenaccesstocreatenewDAOs thatarecompatiblewithmainstreamblockchainplatformssuch asEthereum[12],Avalanche,Binancesmartchain(BSC),Poly- gon, and Solana. The ample and reliable data collected by SnapshotonDAOcommunitiesmotivateustodevelopthefol- lowingin-depthaswellascomprehensiveresearchsurrounding DAOs. Contributions (Fig. 1):Inthisarticle,wediveintotheDAO projectsthatarecreatedandmanagedonSnapshot.Wedevelop theresearchbygraduallyapproachingtheDAObasicconcept, operating mechanism, and relevant techniques, and analyzing thestatisticaldatacollectedfromSnapshot.Ourworkisthefirst studytostrictlyexplorethefeaturesofDAOs,providingin-time guidance for the following readers. Specifically, we detail our contributions here. 1)AstructuralinvestigationonDAOs(SectionII):Intending tobeacompletestudyfocusedonDAOs,weclearthefogsur- roundingthisfuzzytermbypresentingitsunderlyingstructures (e.g.,components,supportivestandards),coremechanisms,and outstanding instances. In particular, based on extensive inves- tigation, we decouple DAO constructions (e.g., decentralized identifier,utilitytoken,smartcontract,ande-voting)andextract a series of metrics to reflect the features (cf. Section II-B) in DAO’s designs. As tokenization plays an essential role in DAOgovernance(discussedinSectionVI-D),wealsosortout the relevant token standards that are essential to the DAO’s incentive. Further, we provide a short list of tools (Table III) thatwellsupportDAOoperations. 2) A comprehensive exploration on the snapshot launchpad (Section III): We study one mainstream DAO-related gover- nance tool, Snapshot, by summarising the features of involved entities, running mechanisms, typical operations, and voting elacstcejorP Fig.2(a) NumberofregisteredmembersineachconsideredDAOproject. Fig.4(a) NumberofvotesofeachproposalamongallconsideredDAOprojects. Fig.2(b) Up-to-datenumberofDAOprojectskickedoffeverymonth. Fig.4(b) DurationofeachproposalamongallconsideredDAOprojects. Fig.4(e) LanguagesdistributionamongallconsideredDAOprojects. Fig.3(a) Partipantionrateswithdynamicgovernancescale. Fig.3(b) VotingchangeswithdynamicDAOprojectsizes. .tsarfnI Fractionofdifferentblockchainnetworksbeingusedfor Fig.4(g) runningeachconsideredDAOprojects. FractionofdifferentIPFSaddressesbeingusedfordatastorage Fig.4(h) ofeachproposalamongallconsideredDAOprojects. gnitoV-e Fractionofdifferentvotingmechanismsbeingusedfore-votingof Fig.4(f) eachproposalamongallconsideredDAOprojects. Fig.4(c) Votingpatterns(intermsofthenumberofcandidatesandvariances &4(d) ofresults)amongallconsideredDAOprojects. Fig.5 NumberofvotesofeachproposalamongallconsideredDAOprojects. Fig.6 ClusteringamongallconsideredDAOprojects. nekoT egasU FractionoftheusageofprevalentDAOtokensandother Fig.7(a) self-issuedtokensintheSnapshot. Fig.7(b) FractionoftheusagebetweendifferentprevalentDAOtokens. Fig.7(c) Fractionoftheusagebetweendifferentself-issuedDAOtokens. strategies. As of Nov. 20222, Snapshot has registered 11K+ spaces (projects), which covers 95% in-the-wild DAOs. How- ever, many of them are inactive with very few members or proposals. We ignore such projects and put our focus on the influentialones.Thus,wecollectthe581mostprevalentDAO projects that contain a total of 16 246 proposals over the span of past 3 years. In particular, we dive into each project and scrutinize included proposals with basic information, voting strategy,proposalcontent, and votingresults. 3) A solid analysis for collected data from snapshot (Sec- tion IV): Based on extensive investigation and exploration, we structure our experimental results from four aspects that separately interpret the project scale, supporting infrastruc- ture, dependent e-voting schemes, and the operational tokens (cf. Table II). We evaluate each item by diving into multiple subaspects.Accordingly,westudythedetailsofDAOmembers (e.g.,numberofparticipants),basicprojectinformation(project duration, language usage, storage condition, underlying plat- form),andvotingprocess(votingpattern,results,distribution, variances,tokenusage,meaningfulcontexts).Withsubstantive 2Anotabledistinctionbetweentwodates:Nov.2022markstheconclusion oftheexperiments,May2023representsboththecompletionofwritingand thedatetowhichthedatawasupdatedforsubmission. 2816 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 evidence, we conclude seven pieces of home-taking messages centralization, high costs, unsustainable tokenization mecha- (Insights➊ -➐ ingreybannersandthetailoflastpage)ashigh- nisms, and immature supporting technologies (refer to Sec- levelsummaries forinterestedreaders. tion V for more information). All such issues are vital and 4) A series of reasonable analyses and discussions for require much notice. To create a better DAO, we need to ex- buildingbetterDAOs(SectionsV–VII):ExistingDAOfieldsare amine these issues at every level of the DAO and strive for absentofrigorousstudiesthatcandelivereffectiveeducational a healthy approach to distributed governance. This involves guidance. We thus provide our discussions based on previ- developingnewtokenizationmechanismsthatincentivizelong- ous empirical efforts. Specifically, we delineate our analyses term participation, finding a more fair governance structure, from three dimensions. 1) Surrounding existing projects, we and exploring alternative blockchain technologies that address study the compatible tools used for DAOs (e.g., on- and off- securityconcerns. chain voting, compatible coordination tools) that can maxi- mally extend the scope of applicability and usage. 2) Diving II. APPROACHINGDAO into each DAO constructions, we point out several unavoid- This section presents a systematic overview of DAOs. We able drawbacks (e.g., centralization, high cost) that may hin- achievethisbybreakingdowntheintegratedcomponents,iden- der the DAO progress and development. Chasing an optimal tifying key features, reviewing the leading DAOs, and dis- balance-off among all distributed projects should be aligned cussingunderlying EIPstandards and surrounding tools. with concrete requirements. 3) Excavating historical failures and the reality of today’s DAOs (e.g., contract reliance). 4) A. DAO Components Exlopring how our findings can be related/reflected to real- world scenarios, in particular for policy considerations. 5) DAOsconsistofvariousintegratedcomponentsthatworkto- Further providing several promising directions that can be gethertofacilitatedecentralizedgovernance,decision-making, improved in the future to fit our identified four aspects in and management. Welistthemajor components. results (e.g., multi-DAO Collaboration, the incorporation of 1) Smart Contract: A smart contract is a piece of code subDAOs). that securely runs on blockchain nodes at the same time in a Spotlight of our Work? We refer specifically to the depth, decentralized manner. Thinking of it as a black-box, both the scale, and real-time relevance of the empirical data used to inputandoutputareguaranteedsynchronizeduponreachinga analyzeDAOsontheSnapshotplatform.Whilethereareindeed consensus without any assistance of trustworthy third parties. studies that address various aspects of DAOs, our research Smartcontractsareconsideredsuitabletoachieveautonomous differentiatesitselfinseveralkey points. organizing by enabling completed self-execution once the 1) Scale and scope of data: Our study utilizes a uniquely defined condition is triggered by traceable and immutable large dataset from Snapshot, covering over 11,000 DAO transactions. This enables real-time auditing and verification projects.Thisscaleallowsustoprovideamorecompre- (e.g., [13]), hence significantly enhancing the machine- hensiveanddetailedanalysisthanmostexistingstudies. executionsecurity[14].InthecontextofDAOs,smartcontracts 2) Real-time relevance: We ensure that our data reflects are often deployed to create multi-sig wallets for secure asset themostcurrentdevelopmentsintheDAOspace,which reservation and setvoting strategiesforfairgovernance. is critical given the rapid evolution of blockchain tech- 2) On-Chain Identifier: Traditional identifiers that rely on nologies. This aspect of our research offers more up- thirdpartiesarereplaced withdecentralized identifiers(DIDs) to-date insights compared with studies that may rely on [15]whicharenotissued,managed,orcontrolledbyanycentral older data. entity.DIDsareinsteadmanagedbyindividualswhoseprefer- 3) Focus on practical implementation: Our research goes encesfordatastorageplatformsareblockchainsuponapeer-to- beyondtheoreticalframeworkstoexplorepracticalimpli- peer(P2P)network.Bymakinguseofpublickeyinfrastructure cationsandapplications,particularlyhowDAOsoperate (PKI) technology to generate an asymmetric key pair that is on modern blockchain platforms and interact with on- stored on the blockchain, DIDs can achieve globally unique, chain governance mechanisms. secure,andcryptographicallyverifiableauthenticationservices. 4) Comprehensive metrics: We introduce new metrics for TypicalimplementationsofDIDsincludeEthereumaddressand evaluating DAOs that have not been systematically ex- Ethereum name service[16]. plored in previous research, such as the analysis of 3) Off-Chain Snapshot: Snapshot is a technique to record governance structures, tokenomics, and the influence of thein-timestatusofdataataspecificheightofblocks.Sucha technological advancements onDAOoperations. typeoftechniqueisquiteimportantforDAOgovernancewhere Key Takeaways. The rapid growth and widespread adoption all the historical results voted by participants are recorded as of DAOs have brought about significant changes in the way evidence,whichisnecessaryforbothon-andoff-chaingover- organizationsarestructuredandgoverned.Onthepositiveside, nance.Off-chainsignaturesareoftenusedtoadjusttheweights we observe that DAO participation and usage are distributed, of on-chain tokens during the voting process. To achieve a application and proposal topics are diversified, and execution smoothcollaboration,asnapshotofon-chainbalancesandad- and decision-making are automated, which are confirmed by dresseswillbecapturedtodeterminevotingrights,andthenthe our empirical observations. However, on the flip side, DAO participantsofcommunitymemberswillstarttovoteforDAO development faces inevitable challenges, including issues of proposals under the weights. In this way, on-chain transaction WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2817 feesaresignificantlywaived.Notably,thenameoftheSnapshot 1) Operational Mechanism: a)Networkreferstotheunder- platforminthisarticleexactlycomesfromthistechnicalterm. lying blockchain platform on which the DAO operates. Since 4) Stake/Governance Token: The self-controlled, and DAOs rely on self-executing contracts where the terms of the portable DIDs can offer tamper-proof and cryptographically agreementaredirectlywrittenintothecode,thenetworkplaysa secure attestations for on-chain decentralized identity. By crucialroleindeterminingthefunctionalityandshapeofthese raising the burden on each attestation’s provenance and smart contracts; b) Protocol/Field describes the specific usage validity to be securely proved, at the same time, easing the or application of the organization. This can range from areas validation process, DIDs become suitable for implementing such as finance and governance to art and social impact; and wallet services in which stakes and utility tokens can be c)Governance tokenrepresentsthevotingpowerinDAOgov- securely stored. Stakes refer to the tokens that a holder can ernance. Stakeholders can thereby vote on proposals to make deposit in the system. The more stakes a holder provides, decisions and allocateresources. the higher confidence he will have in operating consensus 2) Functional Features: Based on the DAO projects on procedures (e.g., proof-of-stake). In contrast, utility tokens Snapshot, we conclude that a typical lifecycle of DAOs in- are designed to be used for a specific purpose, especially in a cludes the phases of create, propose, vote, and action. Specif- DApp or in a game. They offer users benefits such as access ically, 1) create involves setting up the initial configurations to products and services. Staked tokens and utility tokens, in of the DAO, covering not only the DAO space and related most cases, are separate where the former ensures the normal information but also personal identifiers (e.g., ENS, DiD). 2) operationofsystems,andthelatterisusedforgovernance[17] Propose focuses on drafting, editing, and releasing proposals. and votes in the context of DAO. In this sense, communities Specific requirements will be applied to the proposer, such sometimes equivalently use the name of governance token. as holding enough stakes. Vote calls for feedback and pref- Besides, the staked tokens can be further used to establish an erences from community participants. Stakeholders can vote on-chain reputation, which is primarily to give corresponding for multiple options, mostly just for or against based on their credits toindividuals frequentlyparticipating inDAOs. interest and willingness. Different voting strategies will be 5) Reputation Mechanism: Reputation is a crucial element used to adjust the power of a voter. Further, action is to exe- inmaintainingtrustandpromotingcollaborationwithinDAOs. cute the decisions once reaching an agreement. Although this Itservesasameasureofamember’scontributions,determining phaseiscritical,itcannotbeeffectivelymeasured.Wethereby their level of influence. Members can earn a reputation by ac- omitit. tivelyparticipatingingovernancedecisions,providingliquidity 3) Nonfunctional Features: to a protocol, or contributing to ongoing projects. The more a 1) Permissionless is a key factor to measure decentralized member contributes to a DAO, the higher their reputation will governance due to its dynamic joining/leaving mecha- be. This reputation can be leveraged in various ways, such as nisms. Token holders can make decentralized decision- determining voting power in governance decisions, allocating makingbyvotingonpreferredproposalsandinfluencing rewardsfromtheorganization’streasury,orgrantingaccessto theorganization’s direction. certain resources and privileges. Typically, reputation is quan- 2) Transparency/Immutability means that all transactions tifiedviathenumber ofgovernance tokensheldbyamember. anddecisionswithinaDAOwillbetransparentandcan- 6) Secure e-Voting Scheme: Although traditional e-voting notbetamperedwith,fosteringtrustamongmembersand systemshavebeengrowing,theyarestillsusceptibletomanip- stakeholders. ulation.Oneofthemostcriticalproblemsisitbeingpronetothe 3) Anticensorshipreferstotheabilitytopreventstakehold- Sybil attack [18] where malicious users create false identities ers from censoring the flow of transactions (e.g., OFAC tovote.IntheDAOspace,byusingDIDsandaskingforanon- compliant [20]). chainattestation,theintegrityofthee-votingprocesscouldbe 4) InteroperabilityisthefeaturethatallowsaDAOtointer- improved. Staked tokens and utility tokens, which are bound actandexchangedatawithotherDAOs,enablingseam- with DIDs, are also commonly used in e-voting to represent less integration and collaboration with various ecosys- thevotinginfluence.Basedonourinvestigation,existingDAO tems. votingschemesarebasedonrelativelysimplemechanismssuch 5) Token-BasedIncentives.Token-basedincentivesalignthe asbasic voting,single-choice votingandranked choice voting interests of stakeholders and encourage active participa- [cf. Fig. 4(f)], rather than complicated cryptographic e-Voting tion. Members can earn tokens by contributing to the systems[19]. organization or by staking them to support projects, re- sultinginamoreengaged community. 4) Market Performance: Market performance can be eval- B. DAO Features uated viaquantitative metricsinmultipledimensions. We examine DAOs from four key perspectives: operational 1) Treasury denotes a collective pool of funds owned and mechanism (for underlying foundations/dependencies), func- managed by the organization’s members. Analogous to tionalfeatures(namely,processingphases),nonfunctionalfea- the total value locked (TVL) in decentralized finance tures (a.k.a. advanced properties), and market performance (DeFi)[21],thetreasuryaccruesovertimefrommember (equiv.real-worldimpact).Weadditionallysummariseasmall contributionsandprofitsderivedfromtheorganization’s portionof projectsinTable III. operations. 2818 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 TABLEIII MAINSTREAMDAOSANDTOOLSANDDATAFEEDS OperationalFeatures Functionalities Non-Functionalities MarketPerformance[May2023] DAOs Network Protocol/Field Token stcejorP etaerC esoporP etoV sselnoissimreP ycnerapsnarT pihsrosnecitnA ytilibareporetnI evitnecnI yrusaerT )DSU( sredloH slasoporP setoV Uniswap Ethereum DeFi(DEX) UNI ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 2.7B 363k 124 203.8k BitDAO Ethereum DeFi(DEX) BIT ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 2.7B 18.5k 23 4.8k ENS Ethereum NameService ENS ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 1.1B 64.2k 60 111.8k Gnosis Ethereum DeFi(DEX) GNO ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 1B 363k 124 203.8k dYdX Ethereum DeFi(Lending) DYDX ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 903.5M 36.4k 26 11.1k Stargate.Fin Ethereum Service STG ✓ ✓ ✓ ✓ ✓ n/a ✗ ✓ 374.8M 26.8k 47 2.2M Lido Ethereum DeFi(Lending) LDO ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 352.6M 33.2k 128 42.3k Polkadot Substrate Service DOT ✓ ✓ ✓ ✓ ✓ n/a ✗ ✗ 280.4M 1.3M 363 2.17k Frax.Fin Ethereum Stablecoin FXS ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 271.3M 13.2k 276 9.04k Aragon Ethereum Service ETH ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 199.1M 14.2k 606 1.03k Curve Ethereum Stablecoin CRV ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 148.8M 76.8k 221 2.10k Fei Ethereum Stablecoin TRIBE ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 145.8M 14.3k 161 15.1k Decentraland Polygon NFTs MANA ✓ ✓ ✓ ✓ ✓ n/a ✗ ✗ 138.5M 308.9k 2k 94.7k Radicle Ethereum Service RAD ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 126.4M 6.6k 26 686 Aave Polygon DeFi(Lending) AAVE ✓ ✓ ✓ ✓ ✓ n/a ✗ ✓ 124.9M 155.8k 268 527.1k Compound Ethereum DeFi(Lending) COMP ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 121.5M 208.6k 169 13.4k DXdao Polygon DeFi(DEX) DXD ✓ ✓ ✓ ✓ ✓ n/a ✗ ✓ 117.1M 1.4k 915 2.54k Ribbon Ethereum DeFi(Derivative) RBN ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 116.3M 4.4k 31 4.75k Synthetix Ethereum DeFi(DEX) SNX ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 115.3M 91.5k 569 14.6k MangoDAO Solana DeFi(DEX) MNGO ✓ ✓ ✓ ✓ ✓ n/a ✗ ✓ 102.9M 36k 401 3.83k Gitcoin Ethereum Socialnetwork GTC ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 92.2M 33.7k 144 70.4k Phala Substrate Polka’stestnet PHA ✓ ✓ ✓ ✓ ✓ n/a ✗ ✗ 77.6M 3.1k 24 72 Vesta.Fin Polygon Stablecoin VSTA ✓ ✓ ✓ ✓ ✓ n/a ✗ ✗ 67.4M 256.5k 8 34.7k JPEG’d Ethereum DeFi(Lending) JPEG ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 66M 5.3k 59 2.51k Euler.Fin Ethereum DeFi(Lending) EUL ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 63.5M 2.6k 55 8.27k MeritCircle Solana NFTs MC ✓ ✓ ✓ ✓ ✓ n/a ✗ ✗ 61M 8.9k 26 2.83k SuperRare Ethereum NFTs RARE ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 54.1M 8.7k 17 1.23k KeeperDAO Ethereum DeFi(MEV-extractor) ROOK ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 53.5M 17k 41 1.21k MakerDAO Ethereum Stablecoin MKR ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 49.1M 90.9k n/a n/a UXDProtocol Solana Stablecoin UXP ✓ ✓ ✓ ✓ ✓ n/a ✗ ✗ 49.6M 11.7k 819 3.27k Yearn Ethereum DeFi(Lending) YFI ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 37.9M 54.2k 16 4.84k Balancer Ethereum DeFi(DEX) BAL ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 36.1M 45k 378 82.1k PleasrDAO Ethereum NFTs USDC ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✗ 31.4M 149 54 1.02k Sushiswap Ethereum DeFi(DEX) SUSHI ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 28.7M 109.1k 290 49.2k Pangolin Polygon DeFi(DEX) PNG ✓ ✓ ✓ ✓ ✓ n/a ✗ ✓ 19.2M 32.5k 45 2k 1inch Ethereum DeFi(DEX) 1INCH ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 18.2M 87.5k 22 2.45k Lucidao Polygon Service USDT ✓ ✓ ✓ ✓ ✓ n/a ✗ ✗ 11.8M 1.3k 6 154 Kusama Ethereum Polka’stestnet KSM ✓ ✓ ✓ ✓ ✓ ✗ ✗ ✓ 11.5M 291.3k 863 5.65k Serum Solana DeFi(DEX) SRM ✓ ✓ ✓ ✓ ✓ n/a ✗ ✓ 4.5M 226.1k 52 246 Bifrost Substrate DeFi(Lending) BNC ✓ ✓ ✓ ✓ ✓ n/a ✗ ✓ 4M 84.7k 686 1.53k Projects Field Note Projects Field/Coverage Note daphcnuaL&slooT Aragon Managementtools ,strA,TFN,ediuGnosdrowyeK ,emaG,iFeD,3beW,yrusaerT ,rehcnuaL,lageL,DiD,scitylanA ,noitatupeR,ecnanrevoG,aideM etupsiD,laicoS,erutcurtsarfnI sdeeFataDdetaleR Dune Dataanalytic https://dune.com/home DAOStack Managementtools GraphQL Dataanalytic https://daostack.io Colony Managementtools ColonyAPI Dataanalytic https://colony.io Snapshot Off-chainvotingplatform DexTools Tradingpair https://www.dextools.io Tally On-chainvotingplatform DefiLlama DeFiTVLaggregator https://defillama.com DeepDAO Information/aggregator TokenTerminal Projects,Financialdata https://tokenterminal.com DAOMasters Launcher/Management RootData Fundraising,Investors https://www.rootdata.com DAOlist Information/aggregator CoinMarketCap Projects,Ranking https://coinmarketcap.com Mirror Publishing/Writing Zapper DAOs,NFTs,DeFi https://zapper.xyz/daos GnosisSafe Multisigwallets DappRadar DApps,NFTs,DeFi https://dappradar.com IPFS Storageinfrastructure DexScreener Tradingpair,Price https://dexscreener.com X PublicChannel CoinGecko Project,Price https://www.coingecko.com/ ☞ SourcedatainthispapermainlyreferstoDeepDAO(https://deepdao.io/organizations)[May2023]. ☞ Insights-➏ EthereumDAOs(post-Merge)arecensoredduetotheOFAC-compliantblocks(MEVWatchhttps://www.mevwatch.info). ➐ DeFi-relatedDAOsareincentive-compatibleasstakeholdersaremotivatedtoholdandusetokenstomaximizetheirprofits. 2) Holders represent participants who own governance to- 4) Votes count for the total number of votes (equiv. deci- kens and are therefore eligible to vote on proposals that sions)castbystakeholders. shape the direction of the organization. It can provide 5) ACollectionofLeadingDAOs(UpperTableIII): Based insight into the level of participation and engagement ontheaforementionedmetrics,weinvestigateagroupof(30+) withintheDAO. DAO projects that are currently active and operating in real- 3) Proposals are the specific documents that outline a sug- worldenvironments.Theseselectedprojectsarehighlyinfluen- gested course of action for the organization. These pro- tialwithin their respective communities, as evidenced by their posalscan beputforwardbyany member oftheDAO. marketperformance.It’sworthnotingthatmostoftheselected

WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2819 EIP-721/BEP-721/EIP-1155: Common Interfaces for the non-FT (NFT) [26] and multi-token [27]. Stakes and gover- nance tokens can also include the forms of NFTs [28], being the voting power of e-voting or being the deposit for users to participate in any campaigns of a DAO. BEP means the standards ofBSC (BNBSmartChain). EIP-4824:CommonInterfacesforDAOs[29].Thisstandard aims to establish conventions on matching on- and off-chain representations of membership and proposals for DAOs by using daoURI, an indicative reference inspired by ERC-721 [26], which enhances DAO search, discoverability, legibility, and proposalsimulation. EIP-1202: Common Interfaces for the voting process [30]. The standard implements a range of voting functions (e.g., VoteCast, castVote, MultiBote) and informative functions (voting period, eligibility criteria, weight) to enable on-chain votingas wellastoview votingresultsand setvotingstatus. ERC-779: Common Interfaces for DAOs [31]. Unlike other hard forks that have altered the Ethereum protocol, the DAO Forkisexecutedsolelythroughthealterationofthestateofthe DAO smart contract whereas transaction format, block struc- ture, and protocol were not changed. It is an “irregular state change”thatwastransferredetherbalancesfromthechildDAO contracts intoaspecified account. D. Surrounding Tools Additionally, a wide variety of tools have been proposed to ease the process of joining, launching, and managing a DAO. We list several of them at the bottom of Table III. Besides the launchpads that can manage DAOs, a host of providers intro- duce their services and infrastructure [32] such as token ser- vices(e.g.,MakerDAOformaintainingtheDAIstablecoin),on Fig.2. Projectscale.(a)NumberofmembersindifferentDAOs.(b)DAO andoff-chainvotingtools(Tally,Snapshot),treasuryoversight launchingdates. (TokenTerminal, Zapper), growth products, risk management (Gnosis),taskcollaboration (Mirro,Colony), community plat- DAOs operate on the Ethereum blockchain [alsosupported by forms (MolochDAO, Metagovernance), analytic tools (Dune, Fig. 4(g)] and are classified as belonging to the DeFi track RootData), operational tools (Aragon, DAOstack), wallet ser- [supported by Fig. 2(a)]. Additionally, DAOs are expected to vices (GnosisSafe) andlegal services (LegalDAO). havecertainessentialpropertiessuchaspermissionlessaccess and transparency, buttheotherspossessadditional qualities. III. DIVINGINTOSNAPSHOT Snapshot is an off-chain voting system designed for DAOs C. Supporting Standards createdonmultipleblockchainplatforms.Thesystemhasbeen Recall that the standards referred to in this article are for- widelyadoptedbymanycryptostartupsandcompaniestoassist mattedintechnicaldocumentsdedicatedtoon-chainprogram- insurveyingusers.Eachprojectcancreateproposalsforusers ming.Conventionsareestablishedbyusingthestandardsduring to poll votes by using the staked or governance tokens. All programming without having to reinvent the wheel, making thevotingproceduresareessentiallyfeelessastheoperationis it easier and more efficient for applications and contracts to executedoff-chain,avoidingcostlyon-chainverification.Users interact with each other. Here, we list the relevant standards only need to connect their wallet to the launchpad and allow thatsupportDAOscenarios. theactionofsigning.Besides,theprojects,votingproposalsfor EIP-20/BEP-20: Common Interfaces for fungible-tokens eachproject,andcorrespondingresultsarestoredbasedonthe (FT) [22]. Running the e-voting normally requires stakes and IPFS decentralized storage system [33]. The snapshot thereby utility tokens that typically implement ERC-20 on Ethereum becomesaconvenienttoolforDAOcreatorstoquerythefeed- [22], BRC-20 on Bitcoin [23], [24], BEP-20 on BSC [25], or back from the communities. We provide detailed actions for similarstandards onother blockchain platforms. each party. 2820 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 1) DAO creator: DAO creators are those companies or guidelines established by the community manager. For projects that aim to use Snapshot. The creator needs to example, in the ENS community, a user must possess at hold a valid ENS domain and register his project on least10000ENStokenstobeeligibletocreateproposals. the Snapshot launchpad by creating a profile with in- Upon fulfilling these criteria, the proposer may proceed putsofdetailedinformationsuchasprojectname,about, to draft the proposal by specifying the content, options, website, symbol, service, network (equiv. blockchain andvotingrules,aswellassettingthestartandenddates. platforms) and contacts such as Twitter, Github and 3) Poll/Vote: The voting process is open for the commu- CoinGecko. nity only if a user has governance tokens. Every project 2) Poll proposer:Theycancreatetheirproposalsforaspe- has its unique governance tokens where a user can even cific project if he holds a sufficient amount of relevant trade (buy/sell/exchange) them on secondary markets. governancetokens.Inmanycases,pollproposersarethe The voting process is designed in a clean and simple DAO-creatingteammembersastheyhaveenoughstaked style:connecttothewallet,selectoptions,andsignwith tokens and motivations toimprove theprotocol. signatures. Users can view their options, voting power, 3) Users: Users can vote for each proposal based on their andsnapshottimeforeachsubmissionofvoting.Allthe preferences.Allparticipantsneedtohavevalidaccounts dataisobtained fromthesnapshot. with staked tokens for corresponding platforms, such as anEthereumaddressorashortnameregisteredonENS. C. Voting Strategy Users can add a record on accounts to allow votes to be viewable attheconnected addresses. As the most essential part of profit distribution, different strategies provide a series of methods of calculating voting power. The strategy in Snapshot is essentially a JavaScript A. Running Mechanism function.Userscancombineatmost8strategiesoneverysingle TheSnapshotprojectrootsinthetechniqueofsnapshot.The proposal while voting power is cumulative. Meanwhile, users snapshottechniqueistorecordthein-timetoken-holdingstatus canwritecustomizedstrategiesaccordingtotheirrequirements. ofallaccountsandwalletson-chainataspecificblockheight.It Atthetimeofwriting,Snapshothas350+votingstrategiesand actsasthewayofacamera,takingphotosoftheentirepicture ERC20-balance-of is the most adopted strategy. We list their atthemoment.Inthisway,astakeholdercanlearninformation strategies. likewhohasthetoken,howmanytokenstheyhave,etc.Owing 1) Delegated voting: The voting power is based on a del- to the benefits of transparency and traceability, the technique egation strategy. Only permitted stakeholders have valid has been applied to many crypto-events, such as airdrops for impactsonthevoting process. incentivedistributionandcompensationforusersafterhacking 2) Weighted voting: The voting power can be calculated or attacks. Accordingly, the Snapshot project leverages such either by the single weight strategy (one-coin-one-vote) technology to solve the problem existing in the voting pro- or a quadratic strategy. The quadratic strategy weakens cesses. It can intercept the historical data at a certain block thesignificantinfluenceofrichstakeholders,diminishing heightandtheassociatedholdingstatus(e.g.,accounts,tokens, thegap across differentindividuals. NFTs)ofacertaintypeoftoken.Basedonthesedata,thevoting 3) Whitelist voting:Thepermittedstakeholderswhoareon weights can be reasonably assigned to individual community thewhitelistareallowedtovote.Thewhitelistmayeither members aligned withdifferentrules. getupdated manually orby certainrules. 4) NFTvoting:VotingbyusingNFTneedstobecompatible withERC-721 orERC-1155 basedstrategies. B. Typical Operations Based on different roles, we capture three main types of IV. EXPERIMENTSANDRESULTS operations. Notably, operations on Snapshot are aligned with theDAOdeployed onother launchpads. This section provides our experiments and corresponding 1) Creating spaces: If a project aims to introduce decen- results.Wedetailour methods as follows. tralized governance into the project, they can create a Measurement Establishment: Our experiment consists of Space in Snapshot for users to propose proposals and three steps. First, we develop a crawling script, which is perform voting processes. As discussed, a distributed deployed on AWS EC2 cloud server (m6i.32xlarge) with 128- identifierisrequiredbeforetheapplication.Thisidentifier vCPU and 512GiB-memory, to capture all data from the isusedtoconnectthecreateduniqueprojectprofile.The Snapshot platform. The script is designed to collect all the community(equiv.Space)iscreatedoncethebasicinfor- informationpresentedontheSnapshotmainpageandsubpages mationisfullyfulfilled.Importantly,settingthecommu- created by DAO creators, including numerical values (such nity’sdistributionstrategy(a.k.a.,Strategy).Itiswritten as voting results and participation scales) and context-aware aJavascriptfunctionthatcanbeusedtoadjusttheweight strings (such as language usage and topic classification). All ofimpact. data will be compiled into a final CSV document. Second, we 2) Proposing proposals: To submit a proposal within analyzethedatausingPythonandgeneratecorrespondingvisu- the community, a member must first comply with the alizations.Duringthisanalysis,wesort,cleanup,andclassify WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2821 the metadata to obtain meaningful results. Finally, we present our findings andprovide derived insights. OverallStatistics:Ourstudyanalyzes16246proposalsfrom 581 prominent DAO projects, utilizing data collected from Snapshot over a 3-year period starting from its inception in August2020untilNovember2022.Thiscomprehensivedataset includes essential data fields such as project title, number of members, proposal title, status, network, IPFS address, voting strategy, project start/end dates, block snapshots, result name, result outcome, proposal content, and number of votes. For clarityinpresentation,ourstatisticalresultsarecategorizedinto fourkeyareas:project scale,infrastructure,e-voting schemes, and token usage. A high-level overview of our findings is systematicallyoutlined inTable II. A. Project Participation Scale This section describes the scale of the considered DAO projectsinviewofitsparticipatingmembers,votedistributions for proposals, launching dates, active proposal duration, and language distribution.Weprovide detailsofeach item. B. Project Scale Fig. 2(a) illustrates that the top DAO projects achieve a scale of six orders of magnitude (millions) in terms of the totalnumberofmembersweightedbytheirrespectiveproposal counts, with the two most prominent projects, PANCAKESWAP Fig.3. Governancedynamics.(a)Governancescale.(b)DAOprojectsize. andAAVE,surpassing7M.Notethatthe“others”barrepresents the aggregate weighted scale of all DAO projects ranked 16th andbeyond.Thisweightingbyproposalcountunderscoresthe andFrench(1.4%).Alltherestofthelanguagesarecategorized activeparticipationandengagementwithinthesecommunities in “others”, accounting for 12.9% of the proposals. The usage and highlights a Pareto principle-like distribution [34], where of languages can indirectly reflect the nationality distribution the majority of DAO activity is concentrated in a small subset ofparticipating members. of the most prominent projects. When diving into each DAO proposal, it can be found from Fig. 4(a) that the fraction of C. Project Scale Change havingover100votesandhavinglessthan10votescomefirst andsecond,respectively.ThisalsoindicatestheParetoprinciple We further explore how governance models scale changes isbeingcompliedwithinthesensethatahugenumberofvotes withthesizeof theDAOand itstreasury(Fig.3). areaggregated toasmallportionofproposals,whilethereare Fig. 3(a) reveals significant variability in participation rates stillasignificantnumberofproposalsthataremarginalizedby across DAO projects, as indicated by the wide interquartile thecommunity. ranges(IQRs)forsomeprojects.Whilethemeanparticipation Fig. 2(b) shows that the concept of DAO appears to be ac- rate(representedbythereddots)highlightstheoverallengage- cepted and realized by a broader public since Q3 2020 (align ment level, the presence of numerous outliers (black circles) with [32]). From then on, the Web3 supporters kept drawing suggests that certain proposals either attracted unusually high traffictotheDAOcommunity.Itturnsoutthatapeakarosefrom or low participation. This variability underlines challenges in Nov2021toJan2022inregardtothenumberofprojectsbeing sustainingconsistentengagementacrossgovernancedecisions. kicked off during the period (along with the booming of DeFi Furthermore,theboxplotdemonstratesthatsomeDAOsexpe- andNFT).Itcanalsobefoundthattheaveragemonthlynumber riencecentralizedparticipationpatterns,whichcouldhinderthe ofnewprojectsismuchhigherthanthatbeforethepeak,which democratic potential ofdecentralized organizations. indicatesamilestoneinthedevelopmentofDAOcommunities. Fig. 3(b) on the right, illustrating the relationship between AccordingtoFig.4(b),thedurationofeachproposalisfoundto the size of DAOs (number of members) and governance par- bewithinaweek,whichismatchedwiththedurationofmany ticipation (number of votes), shows an overall trend of in- real-world election campaigns. Fig. 4(e) shows that English is creasing votes with larger membership bases. The exponential themostpopularlanguageusedintheproposals,accountingfor trendlinesuggeststhatgovernanceparticipationscalesnonlin- 75.1%oftheproposalsamongallconsideredDAOprojectsin early,withlargerDAOsexperiencingdisproportionatelyhigher ourcollection.Chinesecomessecondwithafractionof4.3%, governance activity. However, this pattern may also reflect in- followed by Germany (2.9%), Korean (1.8%), Italian (1.5%), creasingcentralizationofvotingpowerorvaryingengagement 2822 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 Fig.4. Snapshotempiricalresultsinmultidimensions.(a)Numberofvotesofeachproposal.(b)Durationofeachproposal.(c)Proposalvotingvariances. (d)Projectvotingvariances.(e)Languagedistribution.(f)Fractionofe-votingschemes.(g)Fractionofblockchains.(h)FractionofIPFSstorage. strategies employed by larger DAOs. The absence of extreme storage as illustrated in Fig. 4(h), the “#bafkrei” is used the outliers improves the clarity of this relationship, reinforcing mostoften,accountingfor23.7%amongalltheproposals.This the significance of size as a factor influencing governance impliesthat76.3%oftheDAOproposals(startingwith“#QM”) dynamics. arestillusingtoolssuchasnmkr.ioorothermintingplatforms Ourobservationsemphasizetheneedforrobustmechanisms thatusetheoutdatedversionofcontentidentifier(CIDv0[36], tofosterequitableparticipationingovernanceprocesses,espe- Base58) which is more expensive and less effective for IPFS ciallyasDAOsscaleinsize.Additionally,strategiestomitigate data storage. This deserves a caution that there are lack of voter apathy and ensure fair representation of diverse member motivation for DAO proposers or developers to upgrade their voicesarecriticalforthelong-termsustainabilityofdecentral- infrastructure,whichmightdegradethecapacityandefficiency izedgovernance. of data storage to the DAO community if IPFS would only completely rootforCIDv1[36]inthefuture. Insight-➊ : The DAO community has experienced sig- nificant growth across numerous countries and regions, Insight-➋ : It is fortunate that the platforms used by maintaininghighusagerates.However,thedistribution existingDAOproposalsandprojectsarediversified.On ofmembersoftenadherestotheParetoPrinciple(the80- the contrary, the insufficient motivation of upgrading 20rule[35]),indicatingaconcentrationthatcouldlead the content identifier version of IPFS data storage may tounintendedcentralization.Thisphenomenonwarrants degrade thecapacity andefficiency ofthecommunity. careful monitoring and intervention to preserve the de- centralized ethos ofDAOs. E. E-Voting Scheme This section describes the number of valid votes and the D. Infrastructure fraction of different e-voting schemes used in the considered ThissectiondescribestheinformationofIPFS network and DAOprojects andproposals,anddifferent voting patterns. blockchainplatforminfrastructurebeingusedbytheconsidered 1) E-Voting Schemes: It is found from Fig. 4(f) that the DAOprojects orproposals,followed bydetails. e-voting schemes can be categorized into the following ranks. ItisfoundfromFig.4(g)thatEthereumMainnet[12]isthe Thesingle-choicevotingisthedominantstrategythataccounts mostpopularblockchainplatformusedbytheconsideredDAO for 83.0% among all reviewed strategies, followed by the ba- projects in our collection, accounting for a fraction of 65.4%. sic voting strategy with a fraction of 7.2%. Conversely, the BinanceSmartChainMainnetcomessecondwithafractionof weightedvoting(5.2%),approvalvoting(1.7%),quadraticvot- 14.3%, followed by Polygon Mainnet (8.9%), Fantom Opera ing (1.6%), and ranked choice voting (1.3%) strategies are (3.3%), Arbitrum One (1.4%), and Gnosis Chain (1.4%). All rarely adopted by participating members in comparison. The theotherplatformsarecategorizedin“Others”,accountingfor resultsdemonstratethatsingle-choicevotingisthemostpopular 5.2% of the projects. In regard to the decentralized IPFS data strategy.ThisindicatesthatDAOusersstillprefertoadoptthe WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2823 Fig.6. ClusteringamongallconsideredDAOprojects. Fig.5. Votingpatternswithcandidatesize. Insight-➌ : Current e-voting patterns and results in simplest way of polling. Although an intuitive concern comes many DAO projects illustrate both the decentralization that the single-choice voting and basic voting strategies may and democratic nature of DAO communities. However, result in Matthew effect [37] in vote distribution, the results these systems often struggle to reach consensus effec- show that most proposal advisors ignore such drawbacks in tively,contrastingwiththeefficiencyseenintraditional practice. e-votingsystems.Thischallengehighlightsafundamen- 2) Voting Patterns: Fig. 5 reveals that binary voting is talweaknessinflatorganizationalstructures,wherethe the most prevalent pattern in our dataset, with over 10,000 trade-offbetweenflatandhierarchicalmodelscontinues proposals,followedbyternaryandquaternaryvotingpatterns. tobeasubjectofdebate. Conversely, Fig. 4(c) and 4(d) analyze the variances in voting results for individual proposals and projects, respectively, to determine the extent of agreement or opposition within the 3) Clustering the Voting Contexts: The clustering among community.FromFig.4(c),itisobservedthatmorethan60% allconsideredDAOprojectsisinvestigatedinFig.6.Weapply of the proposals exhibit a high variance, exceeding 40, sug- K-means clustering [38] to analyze and categorize different gesting that e-voting within current DAO communities tends DAOprojectsbasedonthetextualfeaturesextractedfromtheir toproducepredominantlyone-sidedoutcomes.However,atthe titles. As the titles are strings, we first preprocess the dataset project level, the incidence of significant variance is consider- bytransformingthetextualdataintonumericalrepresentations, ablylower,atonly9.2%.Balancedresultsconstitute38.5%of such as term frequency-inverse document frequency (TF-IDF) cases, while 52.3% of projects show an average variance be- orwordembeddings. tween10and20,indicatingthatone-sidedvotingoutcomesare To effectively visualize and interpret the resulting clusters, relativelyrarewhenviewedacrossmultipleprojects.Typically, wethenutilizetwowidely-useddimensionalityreductiontech- althoughindividualprojectsmayexperienceseveralone-sided niques, namely principal component analysis (PCA) [39] and votinginstances,themajorityaremorelikelytoyieldbalanced t-distributedStochasticneighborembedding(t-SNE)[40].PCA outcomes. is a linear technique that identifies directions of maximum Here, we try to explain the reason that causes the variance variance, transforming the high-dimensional data into a one- differences between proposal- and project-level voting results. dimensionalrepresentation(pca-one).Thisprovidesanintuitive As observed in Fig. 5, most of the voting results are binary- visualization and interpretation of the resulting clusters, pre- based patterns, whose corresponding variances are naturally servingasmuchoftheoriginalvarianceaspossible.Conversely, very large. This will significantly increase the result (value) t-SNEisanonlineartechniquethatpreservesthelocalstructure of proposal-level variances as each proposal is merely estab- of the data, capturing complex patterns and relationships. We lished on top of one voting pattern. In contrast, the results uset-SNEtogenerate a two-dimensional representation (tsne- in project-level variances are relatively balanced because each 2-D-one)oftheDAOprojecttitles,enablingadetailedexamina- project contains a series of proposals that may moderate the tionofclusters,substructures,andintricaterelationshipsamong extreme value caused by binary results. In our view, one- theprojects.Byconductingboth,weobtainaricherunderstand- sided results do not necessarily mean “bad”, which instead ingoftheunderlyingpatternsandrelationshipsamongdifferent indicate that DAO members tend to make an instant decision DAOsbased ontheirtitles. withoutsignificantdebates.AbalancedresultshowsthatDAO Weclustertheprojectsinto10labels(cf.Fig.6)alongwith communities are difficult to reach an agreement among the briefsummaries aspresented below. participants. However, on the flip side, this exactly reflects Label 0 (cid:2) <Protocol Upgrades and Implementations>: the so-claimed properties of decentralization or democracy. This category focuses on proposals related to upgrades, im- Controversial arguments indicate that defining what is a nor- plementations, and enhancements of decentralized protocols mal or healthy voting result is complicated in an unclear andplatforms.Topicsincludenetworkupgrades,smartcontract context. implementations, andconsensus mechanism improvements. 2824 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 Label1(cid:2)<GovernanceandDecision-Making>:Thiscate- technical improvements. Label 4 (cid:2) encompasses marketing- goryprimarilycoversvariousproposalsanddiscussionsrelated relatedeventsandiscloselylinkedwithfinancialmanagement to governance, management, and decision-making processes representedbyLabel5(cid:2)andassetmanagementrepresentedby withinDAOsandotherdecentralizedorganizations.Topicsin- Label 7 (cid:2). Label 8 (cid:2), concentrating on contest events, stems cludevotingsystems,governancestructure,andvariousaspects fromtheinitiativesandactivationcharacterizedbyLabel9(cid:2). ofadministration. Simultaneously, both Label 8 (cid:2) and Label 9 (cid:2) intersect with Label2(cid:2)<Tokenomics,Staking,andRewards>:Thiscate- Label 2 (cid:2), highlighting that the activation and continuation goryisfocusedontokenomics,staking,rewards,andincentives of individuals or projects within DAOs necessitate extensive fordecentralizedplatformsandprotocols.Discussionsandpro- discussions about adopting appropriate token incentives. Con- posals revolve around token distribution, staking mechanisms, versely, Label 1 (cid:2) and Label 2 (cid:2), positioned at the center, yield farming, liquidity provision, and other related financial validatethatthegovernanceandtokenomicscomponentsform aspects. the core of DAOs, aligning with the introduction presented in Label 3 (cid:2) <Development and Technical Improvements>: Section II-A. This category deals with discussions and proposals related to thedevelopment,improvement,andmaintenanceofdecentral- Insight-➍ : DAOs exhibit a broad range of voting izedplatforms,protocols,andapplications.Topicsincludetech- contexts, covering topics from budget allocations and nicalimprovements,bugfixes,newfeatures,andotheraspects project funding to community events and hiring deci- ofsoftwaredevelopment. sions. This diversity showcases the potential for de- Label 4 (cid:2) <Marketing, Branding, and Community centralized governance to empower communities and Building>: This category covers marketing, branding, and driveinnovationacrossvariousdomains.However,chal- community-building efforts within the decentralized ecosys- lenges such as voter apathy and the concentration of tem.Topicsincludecommunityengagement,socialmediapres- power among a few token holders highlight the need ence,promotionalcampaigns,partnerships,andcollaborations for more robust, inclusive, and accessible governance toincreasevisibilityandadoption. mechanisms that encourage broader participation and Label 5 (cid:2) <Budgets, Funding, and Financial Man- ensure asustainable futureforDAOs. agement>: This category focuses on various budgets, fund- ing, and financial aspects related to DAOs and projects. Discussions and proposals revolve around allocating re- F. DAO Tokens Usage sources, managing expenses, funding proposals, and financial matters. ThissectiondescribestheusageofdifferentDAOtokensused Label 6 (cid:2) <Project-related Requests and Resources>: intheconsidered DAOprojects orproposals. This category encompasses project-related requests, includ- Fig. 7(a) reveals that 97.1% of the DAO projects use self- ing requests for resources, support, or collaboration from issued (equiv. customized) tokens or minor tokens, while only communities. Topics include project funding, hiring, develop- 2.9% of the DAO projects use the mainstream tokens in- mentservices,andresourcesneededtomoveaprojectforward. cluding USDT (54.2%), ETH (24.6%), USDC (18.3%), and Label7(cid:2)<AssetManagementandAcquisitions>:Thiscat- ENS (2.9%), as shown in Fig. 7(b). The results reveal a risk egorydealswithassetmanagement,acquisitions,andpurchases of the current usage of tokens in DAO spaces. The major- withinthedecentralizedecosystem.Topicsincludebuyingand ity stays on using self-issued tokens or minor tokens which selling NFTs, real estate in virtual worlds, and other digital are much less stable and have much fewer merits than the assets, as well as decisions regarding strategic investments or prevalent tokens. Unhealthy opportunistic behaviors could be acquisitions. apparently realized, which is adverse to leveraging smooth Label 8 (cid:2) <Contests, Competitions, and Events>: This and efficient governance. Across the DAOs using self-issued category is focused on contests, competitions, programs, and tokens, the top 3 are STALK, HUWA, and PEOPLE whereas events within the decentralized ecosystem. Discussions and HUWA is tailored specifically to internet memes compared proposals revolve around voting on the outcomes of various withSTALKfacilitatingafiatstablecoinprotocolandPEOPLE competitions, participating in events or programs, and other aiming to develop the subDAOs, as shown in Fig. 7(c). This community engagement activities. impliestheimmaturityofDAOcommunitiesandneedsfurther Label 9 (cid:2) <Activation and Continuation>: This category improvement. coverstheactivationandcontinuationofindividualsorprojects Another interesting observation is that most customized to- within decentralized organizations. Topics include activating kens [over 75% among “Others” in Fig. 7(a)] are minted on new members, continuing or ending ongoing initiatives, ad- the top of Ethereum ecosystems, which means they are in- justing reward structures, and other decisions related to the triguinglydesignedinformsofERC-20tokensthatareclosely management ofhuman resources andprojects. reliedonthedevelopmentofEthereumplatforms.Similarly,the 4) Relations Between Labels: Evidence of correspondence rest of the customized tokens are created on other mainstream betweenthelabeldescriptionsandtheclusteringoutcomescan public chains, such as BSC and Avalanche. Such situations be observed in Fig. 6 through select examples. Label 0 (cid:2) indicate a potential threat of implicit centralization caused by adjacent to Label 3 (cid:2) signifies the close relationship between oligopolisticblockchainorganizationsthathavetakenthefirst- executing protocol upgrades and specific development and mover advantages.

WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2825 a phenomenon raises questions about whether we can truly achievethepromiseofdecentralizedgovernanceinthelongrun. To avoid centralization, DAOs could accordingly prioritize diversity,decentralizedecision-making,avoidconcentrationof assets, embrace transparency, and foster community. Having a diverse group of participants from different backgrounds and expertise can prevent power from being concentrated in the hands of a few. Decentralizing decision-making by allowing all members to participate in governance and voting, through mechanisms like quadratic voting and delegation, can prevent the decision-making process from being controlled by a small group. Avoiding the concentration of assets in a single wallet orexchangecanreducetheriskofasinglepointoffailure.Em- bracing transparency by making all decisions and transactions publicly visible can prevent any hidden centralization from occurring.Additionally,fosteringasenseofcommunityamong members,despiteitbeingprettydifficult,mayhelpensurethat everyone feelsinvested inthesuccess oftheorganization. B. Disunity and Fairness DAO communities come across disagreements much more often than a traditional organization does [cf. Fig. 7(c), also mentioned[8],[10]].Whilethisreflectsthedemocraticnature of DAOs, it also highlights the potential for disagreements to divide the community. A disagreement can arise over a wide rangeofissuessuchasstrategicdirection,resourceallocation, or operational procedures. If left unresolved, a disagreement Fig.7. Distributionofthetokenusage.(a)Fractionofalltokens(general). can escalate and lead to the formation of factions within the (b)Fractionofprevalenttokens.(c)Fractionofself-issuedtokens. community. These factions may then compete against each otherforpower,whichcanunderminethedecentralizednature oftheDAO. Insight-➎ : Unhealthy opportunistic behaviors are still It’s important to have effective mechanisms in place to re- commoninDAOcommunitiesinthesensethatthema- solvedisagreementsinafairandtransparentmanner.DAOscan jority of the projects rather relies on self-issued tokens consider implementing dispute resolution protocols or media- thantheapparentlymorevaluableandstablemainstream tionprocessestoaddressdisagreementsandpreventthemfrom tokens suchasUSDT,ETH,etc. dividing the community. By addressing disagreements proac- tivelyandcollaboratively,DAOscanmaintaintheirdemocratic and decentralized nature while avoiding factionalism and pre- V. DISCUSSIONSONTHREATS serving their collective decision-making power. Additionally, DAO governance should dictate any progress updates for the Thissectionhighlightspotentialchallenges.Theanalysisof projectsourcecodeorotherinitiativesinafullytransparentway threats islargelybased onempiricalevidence gathered. viapubliccommunication channels, e.g.,Discordand Slack. A. Centralization C. Legality The governance in DAOs relies prominently on the posses- It is evident that the majority of successful DAOs operate sion of stakes or utility tokens. Although it is originally ex- within the financial sector [cf. Fig. 2(a), Fig. 7(c), and Fig. 6, pectedtobecoretothedecentralizationinDAOs,highlyactive Insight-➐ ], which poses significant risks from various fronts. groupsofparticipantstendtoaccumulatemajorsharesoftokens Theserisksinclude potentialattacks frommaliciousactors,as [investigated by our results Fig.7(a) and 7(b)], hence breach- well as the threat of being censored by governmental entities ing the decentralization due to the concentration of e-voting (e.g., more than 51% block proposers in Ethereum 2.0 are power. Beyond that, it’s disheartening to observe the growing OFACed by U.S. government, referring to our Insight-➏ ). As centralization of various aspects within DAOs. For instance, a result, smaller organizations may face severe limitations on languageusage[seeFig.4(e)],votingstrategy[Fig.4(f)],plat- theirabilitytooperateeffectively,andinsomecases,theserisks formadoption[Fig.4(g)],andevenstorage[Fig.4(h)]allseem could even leadtotheirdemise. tobefollowingasimilarpathtowardcentralization.Thistrend Properly embracing legal regulation can avoid the above has been discussed in depth in our finding Insight-➊ . Such problems. Laws or regulations about blockchain governance 2826 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 needtobeproperlyestablishedbystandardizingthestructures, F. Contract Reliance processes,developments,andtheuseofblockchainandmaking Most of the DAOs rely prominently on the authenticity and everycomponent(e.g.,DAO)compliantwithlegalregulations validityofthesmartcontractsthatoffertrustlessenvironments. and ethical responsibilities [17]. In particular, after The DAO Thisimpliesthatthevulnerabilityofsmartcontractcodesand hack, DAOs started to be concerned about being legally man- implicit design pitfalls will pose potential threats to running agedwithbettersecurityandprotectioninseveralcountriesand DAOs.Afamoushistoricalexamplecausedbycontractpitfalls regions. is the huge failure of The DAO hack due to a severe bug in its smart contract code [41]. TheDAO raised $150M+ (by D. High Cost ETH) for building a collective investment platform. However, the project crashed shortly afterward due to a severe bug in Running a DAO on-chain can be expensive (Fig. 7), its smart contract code. As a result, a considerable amount of with costs varying based on factors such as the underlying assets was siphoned off and a disruptive hard fork happened blockchainplatform,complexityofsmartcontracts,andtrans- that significantly affected the entire Ethereum blockchain till action volume. These costs are incurred through gas fees paid now [42]. Attacks such as flash loans [43] in DeFi protocols tothenetwork,whicharecollectedbyminersorarbitragebots that exploit the time interval of block confirmation can also and can become expensive in US dollars. Many DAOs create undermine thesustainabilityofDAOcommunities. theirownERC20tokens[Fig.7(c)]touseasgovernancevotes, To prevent a recurrence of such fiascos and stabilize mone- whichalsoincursgasfeeswitheachactiontaken.EvenDAOs tization mechanisms for sustained growth, DAO communities that use stablecoins [Fig. 7(b)] for voting power still need to must dedicate resources to establishing security protocols for purchaseorborrowthecoinsfromexchanges,addingtotheex- code auditing and developing enhanced tools and supportive penses.Additionally,feesfordevelopment,maintenance,audit- infrastructure. Additionally, the creation of robust marketing ing, security assessments, marketing, and community building andproductdesigndepartmentsisessential.Thesedepartments canbedifficulttoquantifyand areexcluded. should develop effective product and content strategies that AreasonablewaytoreducethecostsofoperatingaDAOisto align with the principles of each DAO project. Concurrently, relyonoff-chainorlayer-twotechniquesthatcanexecutemost well-organized and consistent communication plans are cru- operations locally. Snapshot is an off-chain platform designed cial for capturing broader public interest and fostering loyalty tomanageDAOsandenablevotes.Additionally,otheroff-chain withinadecentralized context. tools can be found in Table III to further reduce costs. By leveragingthesetechniques,DAOoperatorscanminimizetheir relianceoncostlyon-chaintransactionsandreducetheiroverall VI. FURTHERACTIONS expenses. In this section, we continue the discussions of previous so- lutions andconduct amoredetailed analysis ofeach category. E. Nonsense Governance Activity A. On Projects Afteranalyzingthevotingcontexts(e.g.,proposaltitlesand topics),wehavefoundthatanonnegligibleproportionofgov- Eachprojectinvolvesbothcompetitionandcooperation,and ernance activities are nonsensical in nature (consistent with a wewilldiscussthemfromthesetwoperspectives. recent report by [8]). Our analysis reveals that a considerable 1) DAO-2-DAO Collaboration: The interaction between number of proposals (approximately 17.7% of all proposals, differentDAOsiscrucial,promptingthedesignofdecentralized raw data of Fig. 6) are completely irrelevant to the project’s negotiation protocols [44]. Governance in DAOs significantly development, and merely consist of inappropriate or offensive hinges on these protocols, with each DAO defining a unique content such as jokes and impolite questions. We think that set of parameters for consensus based on a standardized for- the current ease of proposal creation, which allows anyone to malization of components such as the proposal format. This submit a proposal, has contributed to the prevalence of such standardization not only routinizes the evaluation of propos- nonsensical activitieswithinthegovernance process. als,enhancinginteractionefficiencyforactivitiessuchasjoint Thus, the implementation of more stringent entry require- ventures, token swaps, and shared monetary policies but also ments for proposal creation is necessary, such as mandatory facilitates more complex DAO-to-DAO transactions. Specifi- completionofatutorialongovernance principlesorholdinga cally, a well-crafted formalization advances DAO-2-DAO col- minimumnumberofprojecttokens.Byintroducingsuchmea- laboration toward an interorganizational framework, allowing sures,weexpecttoseeanimprovementintheoverallqualityof proposals from various DAOs to be adapted for a broad array proposalsandareductioninthenumberoffrivolousorfraud- of complex contractual agreements. This extends their func- ulent proposals. In addition, we recommend the establishment tionalityandsupportsamoreintegratedapproachtodecentral- of a mechanism to flag and remove any proposals that violate ized governance, enabling seamless integrations and strategic theplatform’stermsofserviceoraredeemedinappropriateby alliances between DAOs. the community. This could be done through the appointment 2) Learn on SubDAOs – Management/Competition: DAO of community moderators or the development of automated managementhasbeenevolvingtofeatureatreestructureindi- systemstodetectsuchproposals. catingthehierarchyofdifferentDAOswhereonemightbelong WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2827 totheothers.Therewillbenewgroupsofmembersthatoperate 1) Voting Routes: Voting could be conducted through both independently of the group’s inception as DAOs grow. New on-chain or off-chain. The on-chain voting service, such as divisions,teams,focus,andideaswillbebroughtintothecom- Tally[54],hastointroducethetime-lockmechanismtoprovide munity. Rather than trying to house all that activity under one the polling period. Implementing such a mechanism typically roof,SubDAOsareanemergingapproachfordifferentworking relies on the usage of smart contracts. Tally’s voting contains groupstocreatetheirownfoundationandownershipstructure two types of smart contracts: a token contract and a governor [45].AlltheSubDAOstievaluebacktotheoriginatingentity. contract.Meanwhile,themultisigwallet(e.g.,GnosisSafe[55]) At the same time, one thing to be noted is the competition is necessary for managing the deployed assets. However, on- among different subDAOs within the same domain. Multiple chainvotingconfrontsthedisadvantagesofcostlyanddelayed DAOparticipantswillcompeteforonegoalsetbyitssuperior confirmation,significantlydecreasingthewillingnessofpartic- nodes.Balanced-offgamesamongsubDAOsshouldbefurther ipationofusers.Incontrast,Snapshotisanoff-chainvotingtool considered forsuchscenarios. thatremovestheexpensiveconsumptionofon-chaininteractive operations.ThenumberofcreatedDAOspacesintwoplatforms indicates that users are much more willing to participate in a B. On Infrastructure gas-freeplatform. In addition to guaranteeing the secure operation of the core 2) StrategyDesign: ThedesignofvotingstrategiesinDAOs blockchain,awell-developedinfrastructureandarangeofuse- plays a crucial role in ensuring effective decision-making and fulapplicationsarecrucialforpromotingthewidespreadadop- fosteringuserparticipation.Thesestrategiesshouldstrikeabal- tionofDAOs. ancebetweensecurity,efficiency,andinclusiveness,accommo- 1) DAOStacksandTools: Asagenericterm,theDAOspace datingvariousvotingtools,applications,andregulatoryrequire- hasincludedavarietyofprojectsthatcovermanycomponents ments.On-chainandoff-chainvotingmethodscanbecombined andfields.Wecouldsketcharelativelyclearpicturebylearning to create hybrid strategies, leveraging the strengths of each from its “stack” (Daostack [46]). The foundation is the basic approach. For instance, off-chain voting through tools such as andbackendsoftwaremodulessuchasvotingmechanisms(as Snapshotcanbeemployedforpreliminaryorlesscriticaldeci- discussed before) for decentralized governance. On top of it, sions,allowingforamoreagileandgas-freevotingprocess.On a library layer used to build models for back ends is estab- the other hand, on-chain voting, such as the Tally mechanism, lished (e.g., Arc.js [47]). Also, a caching layer is needed for canbereserved formorecriticaldecisions,wherethesecurity collecting and structuring data (e.g., The Graph [48]). On the andimmutabilityprovidedbyblockchaintechnologyareessen- top,theapplication layer isdesigned forDAOuserstodeploy tial.Anotherdimensiontoconsiderinvotingstrategydesignis orparticipateinDAOs(Aragon[49]).Inaddition,avarietyof theDAO-to-DAOvotingmechanism,whereoneDAOcanpar- widely used coordination tools, such as Twitter, Discord, and ticipate in the decision-making process of another DAO. This Github, play a role in supporting and facilitating DAO games can promote cross-DAO collaboration and resource sharing fromanexternal perspective. [56], fostering synergies within the decentralized ecosystem. 2) Applications via DAO: DAOs have been considered as VotinginSubDAOscanalsobeutilizedtofacilitatethedelega- one of the biggest innovations in the Blockchain ecosystems tionofdecision-makingpowertospecializedgroups,enabling [50]. Therein, crowdfunding is one of the prime applications efficiency ingovernance. where DAO plays a vital role. For instance, ConstitutionDAO successfully pulled together $47 million worth of ether in a D. On Tokenization week to try to buy a first-edition copy of the U.S. Consti- tution at a Sotheby’s auction [51]. Besides, DAO has been Tokenization forms the foundation of the blockchain econ- involvedindemocratizingtheMetaverseecosystembyoffering omyandincentivemechanisms.Achievingsustainabilityanda contributions to decentralized infrastructure [52]. In addition, healthy Web3 ecosystem requires buildingontokenization. the paradigm of DAO paradigm is also applied by NFT-based 1) (Un-)healthyTokenization: Ahealthytokenizationdistri- investmentprojects tocreateandconfirmsharedownership of butionenablesfairnesstopeoplewhoareinvolvedintheDAO assets.TheemergenceofanewgenerationofDappsviaDAO projects.Itmeansthatanyonewhoispurchasingthetokencom- in various sectors, e.g., supply chain, finance accounting, IoT, petesonthesametermsandissubjectedtothesametokensales andtransportation[53]hasdemonstratedtheinnovationandthe policies.Besides,reasonabletokenusageofDAOssignificantly needforDAOincurrenttechnologytrends.Especially,DAOis impacts project-controlled liquidity [57]. As discussed in the also investigated that it could be promising for e-government previoussection,theimbalancebetweenself-issuedtokensand systems in improving the efficiency and transparency of gov- mainstreamtokensraisestheriskofmanipulationinthemarket. ernment operations. The equality of token usage logic has significant implications. Forexample,smallmarketcapprojectslaunchedatcheapinitial prices usually face the potential abuse of whale and team pur- C. On Voting Strategies chases. Meanwhile, the larger the market capitalization which Twokeyquestionsregarding votingare:how to cast a vote ismanagedbyDAOs,thelesslikelythechancesthatthewhale and how the outcome of the vote impacts decisions. andinsideteamcanpurchasetokensonthemarket.Hence,the 2828 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 1) Evaluation Results: Fig. 8(a) compares the governance structures of DAOs across various platforms by analyzing the distribution of voting systems. Ethereum Mainnet stands out with the highest number of DAO projects, predominantly em- ployingsingle-choicevoting,followedbyothersystemssuchas ranked-choicevotingandapprovalvoting.BinanceSmartChain Mainnet and Polygon Mainnet also host a significant number of DAOs, but with fewer governance structures represented. Platforms such as Fantom Opera, Arbitrum One, and Gnosis Chain have a smaller number of DAOs and exhibit limited diversityinvotingsystems.Thisanalysishighlightsthatsingle- choice voting is the most widely used governance structure, likelyduetoitssimplicityandefficiency.Incontrast,morecom- plex systems (e.g., quadratic voting, weighted voting) are less common,potentiallybecausetheyrequirehighercomputational overhead orintroduce barrierstoparticipation. Fig.8(b)evaluatestheefficiencyindexofDAOsondifferent platforms.The efficiency index iscalculated by (cid:3) (cid:2) log (M(v)+1)·w(v), ifM(v)>0 E= 10 0, ifM(v)=0 v∈V Fig. 8. DAOs over different platforms. (a) Governance structures over differentplatforms.(b)Effciencyindexoverdifferentplatforms. where 1) E istheefficiency index foraplatform. 2) V isthesetofvotingsystems(e.g.,single-choicevoting, balancebetweenthemainstreamtokensandself-issuedtokens approval voting). notonlyreducesthepossibilityofmarketmanipulationbutalso 3) M(v)representsthetotalnumberofmembersparticipat- provides space and finances forthefounding teamofDAOs. inginvotingsystemv ontheplatform. 2) Governance via Tokenization: Effective governance is 4) w(v) is a predefined weight reflecting the relative ef- crucial for aligning the interests of various stakeholders and ficiency of each voting system in terms of the relative ensuringthestabilityoftheecosystem.Thisisparticularlyim- simplicity, speed, and scalability of each voting mech- portantinDAOswhereincentivizingresponsiblebehaviorcan anism (e.g., single-choice voting: 1.0, multiple-choice be challenging. Tokenization offers monetary incentives [58] voting: 0.8, quadratic voting: 0.6, weighted voting: 0.7, tovariousstakeholders,includingtheprojectteam,application conviction voting:0.5). providers, node operators, blockchain users, and regulators. EthereumMainnetachievesthehighestefficiencyindexdue Ensuring equitable distribution of these incentives is crucial. to a combination of large DAO membership and a predomi- A robust governance framework, incorporating principles of nance of simpler voting systems such as single-choice voting. game theory, is essential to foster diverse stakeholder engage- Incontrast,platformssuchasGnosisChainandArbitrumOne ment and ensure fair representation. Additionally, transparent score lower on the efficiency index, reflecting fewer members distributionofon-andoff-chainincentivescanhelpbuildtrust and lessfrequent useofhighlyefficient votingsystems. and cooperation among stakeholders toward achieving com- 2) ChallengesAcrossthePlatforms: Despitetheseinsights, mon goals. At a higher level, maintaining a balance between DAOsfaceseveralchallengesacrossplatforms.Onhighlyuti- mainstream tokens and self-issued tokens can reduce the risk lizedplatformssuchasEthereumMainnet,scalabilityisasig- of market manipulation. Building better governance through nificant concern. As the number of participants increases, en- rational incentives and transparent mechanisms can lead to a suring active participation and avoiding voter apathy becomes Schellingpoint[59],wheredesirablebehaviorsareencouraged, critical. Platforms with smaller DAOs, such as Fantom Opera andfairnessismaintained. and Gnosis Chain, may face challenges in gaining sufficient participation tomake governance meaningful. Furthermore, more complex voting systems, such as VII. EXTENDEDDISCUSSION quadratic or weighted voting, while theoretically offering greater fairness, can introduce barriers to adoption due to A. DAOs Across Different Platforms their computational requirements and the learning curve for As our empirical study focuses on a single platform, we participants.Securityrisks,suchasvotemanipulationorSybil havebroadenedouranalysistoincludeadditionalplatformsby attacks, remain a concern across all platforms, particularly in presentingcomparative results(cf.Fig.8). decentralized environments whereanonymity isvalued. WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2829 B. Insights to Real-World C. Consistent Validity of Insights WeelucidatetheconnectionbetweenourinsightsintheDAO Toensurethevalidityofourfindings,werevisitedtheDAO ecosystemandtheirreal-worldscenarios,especiallyforpolicy ecosystem using up-to-date data (Jan. 2025) fetched from the considerations. same source, the DeepDAO platform. A comparison between 1) Pareto distribution in members and centralization risks: therankingsbasedonthetotalnumberofmembersweightedby In many cases [60], [61], a small group of participants proposalcountsinMay.2023andJan.2025revealssignificant heldsubstantialvotingpower,whichcontributedtogov- consistency. Notably, 7 out of the top 10 projects in the 2023 ernance vulnerabilities and the eventual exploitation of data [cf. Fig. 2(a)], remain in the top 10 in the 2025 rankings, the system. To mitigate centralization risks, policymak- with projects such as PANCAKESWAP, AAVE, and OPTIMISM ers could consider implementing regulations that pro- COLLECTIVEconsistentlyleading.Thisoverlapunderscoresthe mote equitable token distribution and prevent excessive robustnessandlongevityofthetrendsandinsightspresentedin accumulation of voting power by a few individuals or thisarticle. entities. This would help preserve the decentralized na- Although certain shifts in rankings were noted, such as the ture of DAOs and ensure more democratic governance ascension of projects such as DECENTRALAND and ENS in structures. the 2025 dataset, these variations underscore the dynamic and 2) Platform diversification and technological upgrades continuouslyevolvingnatureofDAOs.Theconsistencyofthe impeding the capacity and efficiency:Forexample,out- majorityoftop-rankedprojectsreinforcestherobustnessandre- dated IPFS versions can lead to slower data retrieval liabilityofourconclusions,whiletheobservedchangesfurther and potential security vulnerabilities, affecting the over- illustrate the adaptability and progressive development of the all performance of DAO operations. Encouraging the DAOecosystem.Thesefindingsconfirmtheongoingrelevance adoptionofupdatedandstandardizedtechnologieswithin ofourinsightsandtheirenduringcontributiontounderstanding DAOscan enhance interoperability and security. Policy- theorganizational dynamics of DAOs. makers might consider establishing guidelines or incen- tives for DAOs to regularly update their technological VIII. RELATEDWORK infrastructure,ensuringtheyremainrobustandefficient. This section covers three dimensions of DAO progress: the 3) Decentralized e-voting patterns and organizational evolution of several major DAOs in the industry, formative structures making it hard to reach consensus: For in- researchonDAOs,and relatedworkonWeb3 governance. stance,theUniswapDAOhasexperiencedchallengesin passing proposals due to low voter turnout and the need for a high quorum [62]. Exploring hybrid governance A. DAO Evolution modelsthatcombinedecentralizeddecision-makingwith We callback several milestones in DAO’s history. The first hierarchical elements could improve efficiency without DAO, known as The DAO [65], was established on Ethereum compromising inclusivity. Policymakers might support in 2016, marking the beginning of DAOs on blockchains. Un- research into such models and provide frameworks that fortunately,theprojectwashacked,ultimatelyleadingtoahard allowDAOstoadoptgovernancestructuresbestsuitedto forkoftheEthereumblockchain[42],[41].Afterthissetback, theirspecific needs. DAOs regained popularity with the emergence of MakerDAO 4) Diverse voting contexts and governance challenges: As [66] in 2018. The project introduced an on-chain governance highlightedin[63],asmallnumberoflargestakeholders systemtoproduceadepositedstablecoinprotocol(a.k.a.DAI). have been observed to wield disproportionate influence Then,in2020,asurgeofdecentralizedfinance(DeFi)protocols overdecisionsintheMakerDAOcommunity,raisingcon- [67], known as the DeFi summer, propelled DAOs to new cernsaboutthetruedecentralizationoftheorganization. heights.Theseprotocolsarebuiltontopofvariousblockchain Implementing policies that encourage broader participa- platforms,suchasEthereum,BSC,andAvalanche,andenable tion and prevent power concentration is crucial. This decentralizedfinanceservicessuchasDEXs(Uniswap,dYdX), could include mandating transparent voting processes, lending (Compound, Aave), yield aggregators (Convex), and setting limits on individual voting power, or introducing staking (Lido), among others. Till now, DAOs embraced the mechanisms that incentivize active participation from a conceptofWeb3[1],andtheirdevelopmentbecameintertwined diversemember base. with the surrounding components that make up the decentral- 5) High risks of excessive opportunistic behaviors and ized web. This includes elements such as wallets, smart con- token utilization: As highlighted in [64], some DAOs tracts,variousblockchainplatforms,andevenregulations[68]. haveexperiencedsignificanttokenvaluefluctuationsdue to reliance on self-issued tokens, affecting their oper- B. Formative DAO Studies ational sustainability. Establishing guidelines for token issuance and utilization within DAOs can help mitigate Liu et al. [2] provide an overview of early DAOs by ex- these risks. Policies that promote the use of stable and plaining the definitions, and preliminaries and giving a simple widelyacceptedtokens,orthatrequireadequatebacking taxonomy. Daian shows a series of DAO attack analyses [69], ofself-issuedtokens,couldenhancethefinancialstability [70] from the technical level by diving into the source code. ofDAOsandprotect members fromundue risk. Theypointoutthereasonsforrecursivesendvulnerabilitiesin

2830 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 Ethereumthatcauseamonetaryloss($150M).Robinetal.[9] Data Use Disclaimer: This work does not raise any ethical have investigated three DAO projects (Compound, Uniswap, issues.AllthedatawecrawlfromSnapshotareopen-released and ENS) by empirically analyzing their voting powers and and free to use with CC0 licenses. We strive to maintain the discussing governance. Later, Daian et al. [71] propose a po- accuracy of all data that we crawl from Snapshot and declare tential attack form called Dark DAO, which means a group of thatthedatawillnotbeusedforany commercial purposes. membersformadecentralizedcartelandcanopaquelymanip- Acknowledgment: An abstract of this work has been pub- ulate (e.g., buy) on-chain votes. Yu et al. [3] provide a quick lishedatICBC’23[76].Anearlierversiontitled“AnEmpirical review of existing DAO literature and deliver their responses Study onSnapshot DAOs”isavailable at[77]. by statistically reviewed papers. Feichtinger et al. [8] conduct anempiricalstudyon21DAOstoexplorethehiddenproblems, REFERENCES including high centralization, monetary costs, and pointless [1] Q. Wang, R. Li, Q. Wang, S. Chen, M. Ryan, and T. Hardjono, “Ex- activities.Feichtingeretal.[72]providedasystematicallystudy ploringWeb3fromtheviewofblockchain,”2022,arXiv:2206.08821. [2] L.Liu,S.Zhou,H.Huang,andZ.Zheng,“Fromtechnologytosociety: onsecuritythreatstoDAOs.Sharmaetal.[10]havedeveloped An overview of blockchain-based DAO,” IEEE Open J. Comput. Soc., their research on 10 diverse DAOs to examine their degree of vol.2,pp.204–215,2021. decentralizationandautonomy.Besides,manyresearchersand [3] G. Yu, Q. Wang, T. Bi,S. Chen,andS. Xu, “Leveragingarchitectural approachesinWeb3applications–aDAOperspectivefocused,”inProc. organizationsalsoputtheirfocusonDAObygeneratingin-time IEEEInt.Conf.BlockchainCryptocurrency(ICBC),2023. reports[32],[73]and onlineposts[68]. [4] Y. Faqir-Rhazoui et al., “A Comparative Analysis of the Platforms for Instead of focusing on single attacks or individual projects, DecentralizedAutonomousOrganizationsintheEthereumBlockchain,” J.InternetServicesAppl.,vol.12/1,no.1,pp.1–20,2021. we provide a comprehensive, data-driven analysis of DAO [5] S.Wang,W.Ding,J.Li,andother,“Decentralizedautonomousorgani- projects hosted on Snapshot. This study goes beyond surface- zations:Concept,model,andapplications,”IEEETrans.Comput.Social levelobservations,offeringuniqueinsightsintomemberpartic- Syst.(TCSS),vol.6,no.5,pp.870–878,Oct.2019. [6] E. Bischof et al., “Longevity foundation: Perspective on decentralized ipation,votingdynamics, andgovernance challenges. autonomous organization for special-purpose financing,” IEEE Access, vol.10,pp.33048–33058,2022. [7] I. Mehdi et al., “Data Centric DAO: When blockchain reigns C. Governance in Web3 over the Cloud,” in Proc. IEEE Int. IOT, Electron. Mechatron. Governance is the cornerstone of DAOs, but research in Conf. (IEMTRONICS), Piscataway, NJ, USA: IEEE Press, 2022, pp.1–7. this area remains vague. DAOs are rooted in blockchain and [8] R.Feichtinger,R.Fritsch,Y.Vonlanthen,andR.Wattenhofer,“Thehid- heavily influenced by on-chain tokens, which means they are denshortcomingsof(d)aos–anempiricalstudyofon-chaingovernance,” affectedbytheunderlyingtechnologyaswellascryptocurrency inProc.Int.Conf.FinancialCryptographyDataSecur.,NewYorkCity, NY,USA:Springer,2023,pp.165–185. regulations. For the former, Kiayias et al. [17] have explored [9] R.Fritsch,M.Müller,andR.Wattenhofer,“Analyzingvotingpowerin the governance process and properties within the blockchain decentralizedgovernance:Whocontrolsdaos?”Blockchain:Res.Appl., context. They use a first principles approach to derive seven vol.5,no.3,2024,Art.no.100208. [10] T.Sharmaetal.,“Unpackinghowdecentralizedautonomousorganiza- fundamental properties and apply them to evaluate existing tions (DAOs) work in practice,” in Proc. IEEE Int. Conf. Blockchain projects.Wangetal.[1]havepartiallyincorporatedthisnotion Cryptocurrency (ICBC), Piscataway, NJ, USA: IEEE Press, 2024, into Web3, emphasizing the importance of DAO governance pp.416–424. [11] Snapshot, 2022. Accessed: Jan. 13, 2024. [Online]. Available: https:// in Web3. Liu et al. [59] have presented a systematic review snapshot.org/#/ of primary studies on governance and provide a qualitative [12] G. Wood et al., “Ethereum: A secure decentralised generalised trans- and quantitative synthesis. Regarding the latter, governance action ledger,” Ethereum Yellow Paper, vol. 151, no. 2014, pp. 1–32, 2014. primarilycomesfromgovernmentsofdifferentcountries,even [13] R. Li et al., “An offline delegatable cryptocurrency system,” in Proc. thoughblockchainisdesignedtobedecentralized.However,it IEEE Int. Conf. Blockchain Cryptocurrency (ICBC), Piscataway, NJ, canstillbeeasilyregulatedbyincreasinglycentralizedminers USA:IEEEPress,2021,pp.1–9. [14] P. Tolmach, Y. Li, S.-W. Lin, Y. Liu, and Z. Li, “A survey of smart or validators. Ethereum (after the Merge) PoS validators who contract formal specification and verification,” ACM Comput. Surv. rely on MEV-boots services may be monitored and regulated (CSUR),vol.54,no.7,pp.1–38,2021. bythegovernment(OFAC-compliant[74]).Additionally,pure [15] R. Drummond, S. Manu, S. Markus, L. Dave, and A. Christopher, “Decentralized identifiers (DIDs) v1.0: Core architecture, data model, cryptocurrencycompaniescanalsobegoverned.Anoteworthy andrepresentations,”2021.Accessed:Jan.13,2024.[Online].Available: example istheUSsanctions againstTornadocash [75]. https://www.w3.org/TR/did-core/ [16] ENS, “Ethereum name service,” 2022. Accessed: May 20, 2024. [On- line].Available:https://ens.domains/ IX. CONCLUSION [17] A. Kiayias and P. Lazos, “SoK: Blockchain governance,” Proc. ACM Conf.Adv.FinancialTechnol.(AFT),2022. We empirically studied the ground truth of the wild DAO [18] J.R.Douceur,“Thesybilattack,”inPeer-to-PeerSystem,P.Druschel, projects collected from Snapshot, the largest off-chain voting F.Kaashoek,andA.Rowstron,Eds.,Berlin,Heidelberg:Springer,2002, platform. Our empirical results discover a series of facts hid- pp.251–260. [19] V. Cortier, D. Galindo, R. Küsters, J. Müller, and T. Truderung, den from the light, covering its unfair distributions on such “SoK: Verifiability notions for e-voting protocols,” in Proc. IEEE as participant regions or language usage, as well as revealing Symp. Secur. Privacy (SP), Piscataway, NJ, USA: IEEE Press, 2016, potential threats of centralization, contract reliance and policy pp.779–798. [20] Q. Wang, S. Fu, S. Chen, and J. Yu, “A first dive into OFAC in DeFi impact.WealsoexploredvariousapproachestoimproveDAO space,” in Proc. Int. Conf. Financial Cryptography Data Secur. (FC), structures.OurworkdelineatesaclearviewofexistingDAOs. NewYorkCity,NY,USA:Springer,2023,pp.133–140. WANGetal.:UNDERSTANDINGDAOS:EMPIRICALSTUDY 2831 [21] E. Jiang et al., “Decentralized finance (DeFi): A survey,” 2023, [49] Aragon, “Aragon platform,” 2022. Accessed: Jan. 13, 2024. [Online]. arXiv:2308.05282. Available:https://aragon.org/ [22] F. Vogelsteller and V. Buterin, “Eip-20: Token standard,” Ethereum [50] Y. Chen, “Blockchain tokens and the potential democratization of Improvement Proposals, 2015. Accessed: May 20, 2024. [Online]. entrepreneurshipandinnovation,”Bus.Horizons,vol.61,no.4,pp.567– Available:https://eips.ethereum.org/EIPS/eip-20 575,2018. [23] Q. Wang and G. Yu, “Understanding BRC-20: Hope or hype,” [51] MacKenzie Sigalos, Constitutiondao crypto investors lose bid to buy 2023. Accessed: May 20, 2024. [Online]. Available: https://ssrn.com/ constitution copy. Accessed: November 26, 2022. [Online]. Available: abstract=4590451 https://www.cnbc.com/2021/11/18/constitutiondao-crypto-investors- [24] N.Li,M.Qi,Q.Wang,andS.Chen,“Bitcoininscriptions:Foundations lose-bid-to-buy-constitution-copy.html and beyond,” in Proc. IEEE Int. Conf. Blockchain Cryptocurrency [52] T.R.Gadekalluetal.,“Blockchainforthemetaverse:Areview,”2022, (ICBC),2024. arXiv:2203.09738. [25] Binance,“Binanceacademy:BEP-20,”2024.Accessed:Jan.13,2024. [53] Y. El Faqir, J. Arroyo, and S. Hassan, “An overview of decentralized [Online].Available:https://academy.binance.com/en/glossary/bep-20 autonomousorganizationsontheblockchain,”inProc.16thInt.Symp. [26] W.Entriken,D.Shirley,J.Evans,andN.Sachs,“Eip-721:Non-fungible OpenCollaboration,2020,pp.1–8. token standard,” Ethereum Improvement Proposals, 2018. Accessed: [54] Tally, “Tally,” 2022. Accessed: May 20, 2024. [Online]. Available: Jan. 13, 2024. [Online]. Available: https://eips.ethereum.org/EIPS/eip- https://www.tally.xyz/ 721 [55] Gnosis, “Gnosis-safe: Trusted platform to manage digital assets on [27] R.Witek,C.Andrew,C.Philippe,T.James,B.Eric,andS.Ronan,“Eip- ethereum,”2022.Accessed:Jan.13,2024.[Online].Available:https:// 1155: Multi token standard,” Ethereum Improvement Proposals, 2018. gnosis-safe.io/ Accessed:Jan.13,2024.[Online].Available:https://eips.ethereum.org/ [56] L. T. Nguyen et al., “Blockchain-empowered trustworthy data shar- EIPS/eip-1155. ing: Fundamentals, applications, and challenges,” ACM Comput. Surv. [28] Q. Wang, R. Li, Q. Wang, and S. Chen, “Non-fungible token (CSUR),pp.1–35,2025. (NFT): Overview, evaluation, opportunities and challenges,” 2021, [57] W. Warren and A. Bandeali, “0x: An open protocol for decentralized arXiv:2105.07447. exchangeontheethereumblockchain,”pp.4–18,2017.Accessed:May [29] T.Joshua,P.Isaac,G.Ido,E.Eyal,Z.Michael,andS.Furter,“Common 20,2024.[Online].Available:https://github.com/0xProject/whitepaper interfacesfordaos,”2022.Accessed:May20,2024.[Online].Available: [58] G. Yu, Q. Wang, C. Sun, L. D. Nguyen, H. Bandara, and S. https://eips.ethereum.org/EIPS/eip-4824 Chen,“MaximizingNFTincentives:Referencesmakeyourich,”2024, [30] Z.Zainan,Evan,andX.Yin,“Erc-1202:Votinginterface[draft],”2018. arXiv:2402.06459. Accessed:May20,2024.[Online].Available:https://eips.ethereum.org/ [59] Y. Liu, Q. Lu, L. Zhu, H.-Y. Paik, and M. Staples, “A systematic EIPS/eip-1202. literature review on blockchain governance,” J. Syst. Softw., vol. 197, [31] D.Casey,“Eip-779:Hardforkmeta:DAOfork,”2017.[Online].Avail- no.1,2022,Art.no.111576. able:https://eips.ethereum.org/EIPS/eip-779 [60] Cryptodose.net,“DAOusecasesandreal-worldapplications.”Accessed: [32] S. Aiden et al., “World economic forum: Decentralized autonomous November 14, 2024. [Online]. Available: https://cryptodose.net/learn/ organizations:Beyondthehype,”2022.Accessed:May20,2024.[On- dao-use-cases/ line]. Available: https://www3.weforum.org/docs/WEF_Decentralized_ [61] Y. Zhang, Q. Wang, S. Chen, and C. Wang, “How to rationally select Autonomous_Organizations_Beyond_the_Hype_2022.pdf yourdelegateeinPoS,”2023,arXiv:2310.08895. [33] J. Benet, “IPFS-content addressed, versioned, p2p file system,” 2014, [62] 101Blockchains.com, “Top DAO use cases and examples.” Accessed: arXiv:1407.3561. November 14, 2024. [Online]. Available: https://101blockchains.com/ [34] Wikiwand, “The pareto principle,” 2022. [Online]. Available: https:// top-dao-use-cases-and-examples/ www.wikiwand.com/en/Pareto_principle [63] C. W. Blog, “Real-world examples of DAOs.” Accessed: November [35] V.Pareto,“Coursd’Économiepolitique,”vol.1,LibrairieDroz,1964. 14,2024.[Online].Available:https://blog.capitalwallet.com/real-world- [36] IPFS,“IPFSDoc:ContentaddressingandCIDs,”2022.[Online].Avail- examples-of-daos/ able:https://docs.ipfs.tech/concepts/content-addressing/#cid-conversion [64] E. C. Blog, “DAO use cases: Exploring potential applications.” Ac- [37] Wikiwand,“Mattheweffect,”2022.Accessed:May20,2024.[Online]. cessed: November 14, 2024.[Online]. Available: https://evacodes.com/ Available:https://www.wikiwand.com/en/Matthew_effect blog/dao-use-cases-exploring-potential-applications [38] S.Lloyd,“LeastsquaresquantizationinPCM,”IEEETrans.Inf.Theory [65] S. Falkon, “The story of the DAO — its history and consequences,” (TIT),vol.28,no.2,pp.129–137,Mar.1982. 2017. [Online]. Available: https://medium.com/swlh/the-story-of-the- [39] H.Hotelling,“Analysisofacomplexofstatisticalvariablesintoprincipal dao-its-history-and-consequences-71e6a8a551ee components,”J.Edu.Psychol.,vol.24,no.6,p.417,1933. [66] MakerDAO, “MakerDAO foundation proposal v2,” 2018. Accessed: [40] L. Van der Maaten and G. Hinton, “Visualizing data using t-SNE.” May 20, 2024. [Online]. Available: https://medium.com/@MakerDAO/ J.Mach.Learn.Res.,vol.9,no.11,pp.1–27,2008. foundation-proposal-v2-f10d8ee5fe8c [41] M. I. Mehar et al., “Understanding a revolutionary and flawed grand [67] S. M. Werner, D. Perez, L. Gudgeon, A. Klages-Mundt, D. Harz, and experiment in blockchain: the DAO attack,” J. Cases Inf. Technol. W. J. Knottenbelt, “SoK: Decentralized finance (DeFi),” in Proc. Int. (JCIT),vol.21,no.1,pp.19–32,2019. Conf.FinancialCryptographyDataSecur.(FC),2022. [42] O. Konashevych, “Takeaways: 5 years after the DAO crisis and [68] Chainlink, “DAOs and the complexities of Web3 governance,” 2022. ethereum hard fork,” 2021. Accessed: May 20, 2024. [Online]. Accessed: May 20, 2024. [Online]. Available: https://blog.chain.link/ Available: https://cointelegraph.com/news/takeaways-5-years-after-the- daos/ dao-crisis-and-ethereum-hard-fork [69] D.Philip,“AnalysisoftheDAOexploit,”2016.Accessed:Oct.10,2024. [43] K. Qin, L. Zhou, B. Livshits, and A. Gervais, “Attacking the DeFi [Online].Available:https://hackingdistributed.com/2016/06/18/analysis- ecosystem with flash loans for fun and profit,” in Proc. Int. Conf. of-the-dao-exploit/ Financial Cryptography Data Secur. (FC), New York City, NY, USA: [70] D. Philip, “Chasing the DAO attacker’s wake,” 2016. Accessed: Oct. Springer,2021,pp.3–32. 10,2024.[Online].Available:https://pdaian.com/blog/chasing-the-dao- [44] C. F. Dagdelen, “D2d: Towards decentralized negotiation protocols,” attackers-wake/ 2021. Accessed: Jan. 13, 2024. [Online]. Available: https:// [71] D.Philip,K.Tyler,M.Ian,andJ.Ari,“On-chainvotebuyingandthe blog.curvelabs.eu/d2d-towards-decentralized-negotiation-protocols- riseofdarkDAOs,”2018.Accessed:Oct.10,2024.[Online].Available: e37d164e91e6 https://hackingdistributed.com/2018/07/02/on-chain-vote-buying/ [45] Coopahtroopa, “How to subdao,” 2021. [Online]. Available: https:// [72] R.Feichtinger,R.Fritsch,L.Heimbach,Y.Vonlanthen,andR.Watten- coopahtroopa.mirror.xyz/7bfK9st2mvhxlla4XKotRjetq5-YhaiwqRwS8D hofer,“SoK:AttacksDAOs,”inProc.Int.Conf.Adv.FinancialTechnol. hkD-o (AFT),2024. [46] Coopahtroopa, “Daostack: An operating system for collective intelli- [73] H. Samer and F. Primavera, “Decentralized autonomous organization,” gence,”2022.[Online].Available:https://medium.com/daostack InternetPolicyReview,2021.Accessed:Oct.10,2024.[Online].Avail- [47] DAOstack,“Arc.js:DAOstackjavascriptclient,”2022.[Online].Avail- able:https://policyreview.info/pdf/policyreview-2021-2-1556.pdf able:https://daostack.github.io/arc.js/ [74] X. Xiong and J. Luo, “Global trends in cryptocurrency regulation: An [48] TheGraph, “The graph: Apis for a vibrant decentralized future,” 2022. overview,”inProc.Int.Conf.Math.Res.BlockchainEcon.(MARBLE), Accessed:May20,2024.[Online].Available:https://thegraph.com/en/ 2024. 2832 IEEETRANSACTIONSONCOMPUTATIONALSOCIALSYSTEMS,VOL.12,NO.5,OCTOBER2025 [75] Q. Wang, et al., “U.S. treasury sanctions notorious virtual currency Caijun Sun received the B.E degree from mixertornadocash,”2022.Accessed:Oct.10,2024.[Online].Available: HangzhouNormalUniversity,Hangzhou,China,in https://home.treasury.gov/news/press-releases/jy0916 2013andthePh.D.degreefromBeijingUniversity [76] Q.Wangetal.,“AfirstlookintoblockchainDAOs,”inProc.IEEEInt. of Posts and Telecommunications, Beijing, China, Conf.BlockchainCryptocurrency(ICBC),Piscataway,NJ,USA:IEEE in2020. Press,2023,pp.1–3. Currently,heisaSeniorSecurityEngineerwith [77] Q. Wang et al., “An empirical study on Snapshot DAOs,” 2022, ZhejiangLab,Hangzhou,China.Hisresearchinter- arXiv:2211.15993. estsincludemalwareanalysisanddatasecurity. QinWang(Member,IEEE)receivedthebachelor’s degree in electronical engineer from the School of ElectronicEngineering,NorthwesternPolytechnical Lam Duc Nguyen (Member, IEEE) received the University,Xi’an,China,in2015,andthemaster’s B.Sc.degreefromHanoiUniversityofScienceand degree in computer science from the Electronic Technology,Hanoi,Vietnam,in2015,andtheM.Sc. Engineering, Beihang University, Beijing, China, degreefromSeoulNationalUniversity,Seoul,South in 2018, and the Ph.D. degree in computer sci- Korea, in 2018, both in computer science, and the ence from the School of Science, Computing and Ph.D. degree in electronic systems from Aalborg EngineeringTechnologies,SwinburneUniversityof University,Aalborg,Denmark,in2021. Technology,Melbourne,Australia,in2022. He is a Research Fellow with CSIRO Data61. He is a Senior Research Scientist with CSIRO His research interests include the intersection of Data61 and an Adjunct Lecturer with University of New South Wales operations research, blockchain, machine learning, (UNSW), Sydney and University of Technology Sydney (UTS), Sydney, andInternetofThings. Australia.HisresearchinterestsincludeemergingWeb3techniques,including Dr.NguyenreceivedtheOutstandingPaperAwardfromWF-IoT2020. NFTs, DAOs, and DeFi, as well as the fundamental aspects of consensus protocols,suchasscalability,security,andprivacy. Shiping Chen (Senior Member, IEEE) received a Bachelor degree from the Harbin University of Science and Technology, Harbin, China, in 1985, Guangsheng Yu (Member, IEEE) received the andMaster’sdegreefromtheChineseAcademyof B.Sc. degree in telecommunication network engi- Sciences(CAS),Shenyang,China,in1990,bothin neering and M.Sc. degree in computer engineer- computer science, and the Ph.D. degree from the ing from the School of Electrical Engineering SchoolofComputerScienceandEngineer,Univer- andTelecommunications,UniversityofNewSouth sityofNewSouthWales(UNSW),Australia,2001. Wales (UNSW), Sydney, Australia, in 2014 and He is a Principal Research Scientist with CSIRO 2015, respectively. He received the Ph.D. degree Data61. He also holds a Conjunct Professor with in cybersecurity from the School of Electrical and University of New South Wales (UNSW), Sydney, DataEngineering,UniversityofTechnologySydney Australia. His research interests include secure data storage & sharing and (UTS),Sydney,in2021. securemultipartycollaboration. Currently,heisaLecturerwithUTS.From2021 Prof.Chenisactivelyinvolvedincomputingresearchcommunitiesthrough to 2024, he was previously a Postdoctoral Research Fellow with CSIRO publications, journal editorships and conference PC services, including In- Data61.Hisresearchinterestsincludecybersecurity,blockchain,andfederated ternational World Wide Web Conference, International Conference on En- learning. terprise Design,Operations,andComputing(EDOC),InternationalConfer- ence on Service-Oriented Computing (ICSOC), and IEEE INTERNATIONAL Yilin Sai (Student Member, IEEE) received the CONFERENCEONWEBSERVICES(ICWS)/IEEEINTERNATIONALCONFERENCE B.Eng. (Hons.) degree in electrical engineering ON SERVICES COMPUTING (SCC)/IEEE INTERNATIONAL CONFERENCE ON (computer)fromtheUniversityofSydney,Sydney, CLOUDCOMPUTING(CLOUD).HeisanIETfellow. Australia,in2017. He is a Senior Software Engineer with CSIRO Data61 and a Postdoctoral Student with Univer- sity of New South Wales (UNSW), Sydney, Aus- tralia.Withexpertiseinfull-stackdevelopment,he specializes in building and delivering innovative, robust, and scalable cloud software platforms and solutions.Hisresearchinterestsincludeblockchain, distributedcomputing,bigdataanalysis,InternetofThings,deeplearningand networksecurity.