Gov-Acc Pipeline.

Proceedings of the 58th Hawaii International Conference on System Sciences | 2025 Is DAO Governance Fostering Democracy? Reviewing Decision-Making in Decentraland AndreaPen˜a-Calvin DavidDuenas-Cid JunaidAhmed UniversidadComplutensedeMadrid KozminskiUniversity KozminskiUniversity andpen03@ucm.es dduenas@kozminski.edu.pl jahmed@kozminski.edu.pl Abstract using digital means, some successful (Borge et al. (2023)), others controversial (Natale and Ballatore This study analyzes voting dynamics and (2014)). A similar pattern can be sought concerning proposal outcomes within Decentraland, a prominent the digitalization of elections, where successful uses Decentralized Autonomous Organization (DAO), by of internet voting, e.g., Estonia (Ehin et al. (2022)), examining its voting behaviors and decision outcomes. go hand-in-hand with other cases where this method We offer insights into how a DAO is employed to wasabandoned,e.g.,NewSouthWales(Haldermanand facilitatedecision-makinganddiscernthenatureofthe Teague(2015))orNorway(Gjøsteen(2016)). issues about which decisions are made. DAOs promise The use of digital methods of digital participation horizontal and democratic decision-making. However, and voting in traditional offline processes (generally our research reveals a high concentration of voting in parallel to conventional offline channels) seeks for poweramongafewmembersdespitethemnotutilizing an increase of convenience for citizens (Licht et al. ittoagreatextent. (2021)), undertheexpectationthatthiswillenlargethe Additionally, we identify the prevailing themes numberofparticipantsengagedinthedecision-making in decision-making processes within the organization process (Borge et al. (2022)) and, eventually, improve through topic modeling. The primary topics identified the quality of democracy. On the other hand, are the effective management and governance of the adopting such technologies in very competitive and platform and community and the platform’s strategic politically charged environments involves several risks growth, with a particular emphasis on wearable andchallengesthatrepresentasignificantlimitingfactor technology. This research addresses fundamental for its wider adoption. Furthermore, an important questions regarding the democratic integrity of DAOs part of the population is still not tech-savvy enough andtheirabilitytoachieveequitablerepresentationand to allow for massive implementation of digital means decision-making. of decision-making. While promising results were obtainedinpilotprojects,weseemtobefarfrombeing Keywords: Blockchain, Decentralized Autonomous abletoadoptthosesystemsatalargescale. Organizations (DAOs), e-Democracy, e-Voting, Online Governance. But how does the use of digital decision-making systems relate to communities or organizations whose 1. Introduction nature is fully digital? Are the same limitations and concerns affecting them? The existing literature The digital transformation introduced changes to describes an interesting reality. Firstly, the lack of almost every dimension of our everyday routines. Our pre-existing forms of governance in many digital experience as individuals is in constant evolution, communities allows for creating their own tailored transformingaspectsrelatedtoourprivacy(Minkkinen governing rules and ad-hoc digital systems (Pitt and et al. (2017)), consumption (Dey et al. (2020)), or Diaconescu (2015) and Schneider et al. (2021)). interaction (Boyd (2010)). The same occurs in our Wikipedia, for example, built its governance system political dimension as citizens: the form we relate combining formal bureaucratic managerial with to public administrations (Gil-Garcia et al. (2020)) or flexible and adhocratic content management structures political life (Vromen (2017)) has evolved towards (Jemielniak(2020)). Secondly, thebaselineknowledge a more digitized interaction, emphasized during the regarding technology use and the digital nature of COVID-19 pandemic. In recent years, we witnessed suchcommunitiesmakedigitalmediatheonlyrealistic several attempts to innovate in social participation optionforengagingthecommunityinitsgovernance. URI: https://hdl.handle.net/10125/109092 978-0-9981331-8-8 Page 2051 (CC BY-NC-ND 4.0) This results in the development of creative digital applications. ADAOcanbedefinedasaself-governed governance models and decision-making systems that organization controlled only and exclusively by an innovate the existing portfolio of democratic means incorruptiblesetofrules,implementedundertheformof (Simon et al. (2017)), introducing, on occasions, asmartcontract(DeFilippietal.(2020)).DAOs,hence, measures to rebalance the distribution of power (e.g., differfromotherorganizations,e. g.,inlackingformal depending on the engagement in the community) or managers, proposing horizontal relations between the mechanisms of voting and decision making (e.g., members, allowing for not long-lasting membership, allowing tracking the results of voting processes on relying on non-hierarchical governance and group real-time) (Goldberg and Scha¨r (2023)). Although it consensus, or using smart contracts to aggregate the seems unrealistic that some of these methods will be votesofmembers(Wright(2020)). transposed to offline decision-making and politically The secure and trust-free assumptions traditionally binding voting processes shortly, the experimental linked to the use of blockchain (Zhang et al. (2019), natureofsomeofthosepracticesisworthobservingto Casey and Vigna (2018)) generated high expectations extract possible outputs on how those features impact of the democratizing power of blockchain-based thequalityofthedemocraticprocessanditsoutcomes. communities and DAOs. For example, Merkle Thispaperanalyzestheuseofonlinevotingmethods (2016) referred to DAO democracy as a new form in Decentraland, a blockchain based Decentralized of democracy which is more stable, less prone to Autonomous Organization (DAO) for entertainment. erratic behavior, able to meet the needs of its citizens, Decentraland claims to be the first decentralized and better using the expertise of all its citizens to metaverse that is built, governed, and owned by its make high-quality decisions. Mainly, expectations users. Unliketraditionalvotingsystems,Decentraland’s about DAOs democratizing power derive from the fact votingisreal-timeandweightedaccordingtomembers’ that they are based on smart contracts and that are symbolicpower,measuredbytheirtokenholdings. The organizations lacking legal, physical, and economic study examines the voting behaviors, discussed issues, constraintsandacentralauthority(Chaoetal.(2022)). and proposals within Decentraland to understand its However, the idea that DAOs will bring a decision-makingprocesses.Usingquantitativemethods, new distribution of power has been seriously it assesses the distribution and use of voting power questioned. Bodo´ et al. (2021) describe how technical among members to determine if the DAO governance decentralization does not necessarily bring social, is democratic. Additionally, the research identifies political, or economic decentralization but can help common themes in decision-making through topic create new centralities of power. Similarly, Cossar modeling. Ultimately, the paper seeks to address et al. (2024) described the democratic limitations questions about the democratic nature of DAOs and of the Proof of Humanity DAO, self-labeled as equitable representation in their decision-making. It “the first democratically governed DAO,” due to its aims to provide insights into DAO governance and its lack of political deliberation, internal polarization, implicationswithindecentralizedecosystems. weak internal legislation, difficult accessibility of the governance platforms, or lack of robustness of voting 2. Theoreticalbackground systems. Liu (2024) reaches similar conclusions, stating that DAOs’ democratic functioning contradicts Decentralization and decentralized organizations its promises by centralizing decision-making power. have been the object of academic attention for several The libertarian aspirations to empower individuals decades, focusing on how decentralization can benefit through decentralization, openness, and freedom (Yee the functioning of organizations (Beckhard (1966)). (2019)) proposed by blockchain technologies are put Theevolutionoftechnologyandsociety, especiallythe intoquestion,alsobythedevelopmentofnewformsof development of networks and blockchain technology, centralizingusesofcryptocurrencies. brought about a renovation in the field, with DAOs Inordertoenrichthisdiscussion,inthisresearch,we resultingfromthisevolution. AsHassanandDeFilippi approach the democratic functioning of Decentraland (2021) put it, the definition of DAO stems from an by analyzing how the features of their governance expansionoftheinitialconceptofDAC(Decentralized system influence the results of their voting processes. Autonomous Corporation), used to refer to the new Specifically, we focus on two elements of the voting corporate governance form, using tokenized tradable process, the participants’ unequal voting power and sharesasameansofprovidingdividendstoshareholders the type of topics discussed. The unequal distribution (Hassan and De Filippi (2021), op. cit.), too of voting power is a generalized feature of DAOs narrowtogathertheparticularitiesofsomeBlockchain governance, following a similar logic to the one Page 2052 proposedbyliquiddemocracyordelegativedemocracy The Decentraland DAO employs Snapshot to store all (Fritsch et al. (2024)). The practical functioning of proposalsandvotessubmittedbythecommunity,along liquid democracy systems has been approached in the with the final results of the proposals. Moreover, the case of the German Pirate Party’s internal democracy DAOoffersauserinterfacewhereuserscreateproposals (Kling et al. (2015)). However, its functioning in and vote, accessible at governance.decentraland.org. DAO voting still lacks systematic studies (Fan et al. Everytimeaproposalisopenedthere,itisautomatically (2023)). Given the topic’s novelty, research still createdinDecentraland’sdedicatedSnapshotspace. did not agree on the impact of unbalanced power For this study, data were gathered exclusively from distribution in the governance of DAOs. While Han Decentraland’sSnapshotspace.Snapshotoffersapublic et al. (2023) conclude that power concentration tends APIthatallowsanyonetoqueryandextractdataabout to reduce its impact when DAOs grow, the works of theDAOsdeployedonthesystem. Byusingit,wehave Fritsch et al. (2024) and Pen˜a-Calvin et al. (2024) extracted all relevant information regarding proposals revealasignificantconcentrationofvotingpowerinthe (titles, votingoptions, andoutcomes), andvotes(voter, governancesystemstheyanalyzed. Combinedwithlow option, and voting power). The data covers all closed participation rates, the previous power concentration proposals from May 11th, 2021 (when Decentraland’s creates a scenario where the influence of active large Snapshotspacewasestablished)toJune4th,2024. We power holders can easily decide voting processes, excludedproposalsthatwerestillopenatthetimeofthe contravening the expected democratizing power of analysis, resulting in a final dataset comprising 2,547 distributedsystems. closed proposals and 166,970 votes. Additionally, the Secondly,thecontentsofvotingprocessesinDAOs number of unique voters was calculated, resulting in a are also uncharted territory due to the difficulties in totalof8,272differentaddressesthatparticipatedinthe systematically analyzing them. Recently, Ziegler et al. DAO governance process. In Decentraland DAO, not (2024) used a Large Language Model to categorize allvoteshavethesameweight. Eachvoterhasitsown proposals reaching the following types of categories: voting power based on the type and number of tokens Treasury and Asset Management, Protocol Risk theyhave,whichmaynotbeconstantacrossproposals. Management,ProtocolFeaturesandUtility,Governance To get an estimate of each voter’s voting power, we Administration and Framework Management, Budget calculate the median voting power of all the votes cast Allocation and Work Management, Partnerships and byavoter. Economic Development, and Miscellaneous. The Governancemetricswereapproachedquantitatively. researchconductedbyCossaretal.(2024)alsocollects Descriptivestatisticsofvotesperproposalandpervoter, some topics discussed in the Proof of Humanity DAO, as well as voting power per proposal and per voter, but given the qualitative nature of their research, the were calculated. The degree of consensus among the resultsdepictanevolutiveanalysisoftopicsratherthan communitywasquantifiedbycomparingthenumberof a comprehensive approach to the contents discussed. votes cast in the winning option of a proposal with the Ma et al. (2024) research describes some relevant voting power, as well as by examining the proportion weaknessesregardingtheinformationavailabilityorthe of unanimous proposals. To assess the distribution of proposal’s description that undermine the democratic votingpoweranditspossibleinequalities,weemployed qualityoftheprocess. theGinicoefficient. Byapproachingtheimpactofpowerdistributionand Finally, we employed exploratory data analytics voting topics in Decentraland, this paper aims to bring (EDA) and topic modeling techniques to analyze the valuable inputs that will continue enriching the current textual data of the proposals’ titles. EDA was used for literature,providingsteppingstonesforfutureresearch. extracting significant insights from textual data (Sahoo et al. (2019)), with a particular focus on elucidating 3. Methodology the DAO resulting decisions. Further, we employed topicmodelingtechniquestogenerateclustersoftopics This research employs a mixed-methods approach, basedontextdatabeingdiscussedatDecentralandasin combining quantitative and qualitative data analysis to Egger(2022).Inparticular,weusedtheLatentDirichlet approachDecentraland’sdecision-making. Theprimary Allocation(LDA)algorithmtogeneratethetopics. data source consists of the proposals submitted and votedon withintheDecentraland DAO,extracted from 4. Dataanalysis Snapshot,apopularoff-chainvotingplatform. Snapshotiscurrentlythemostwidelyusedplatform The data analysis encompasses a total period of 3 forDAOstomakedecisions(Pen˜a-Calvinetal.(2024)). years,fromMay2021toJune2024. Duringthisperiod, Page 2053 2,547 proposals were submitted, 166,970 votes were smallproportionofvoterscanexertadisproportionately cast, and8,272distinctvotersparticipated. Thisshows high influence on decision-making. Figure 1 reflects thatDecentralandisalargeDAOwithhighactivity. these findings through the Lorenz curve. The curve deviates significantly from the line of perfect equality 4.1. Participationandvotingpower (represented by the diagonal black line), underscoring distribution theconsiderableconcentrationofvotingpoweramonga fewvoters. To characterize Decentraland’s voters, it is first necessarytoanalyzetheirparticipationanddistribution 1.0 of voting power. Table 1 indicates that the maximum number of times a voter has participated in the 0.8 governance does not even reach half of the proposals. The mean and average number of times a voter has participatedinproposalsrevealaverylowparticipation 0.6 rate, with voters often participating in less than 1% of proposals,andaverylowvoterturnout. 0.4 Table 1. Voter participation in proposals and 0.2 distribution of voting power. Participation Votingpower 0.0 pervoter pervoter 0.0 0.2 0.4 0.6 0.8 1.0 Accumulated voters Min 1(0.04%) 0% Mean 20(0.79%) 0.01% Median 4(0.16%) 7e-06% Max 1,271(49.90%) 9.16% The estimated voting power distribution among the voters in Table 1 reveals a significant disparity in the influence of each individual. The minimum voting power of a voter is close to zero, while the maximum reaches approximately 9% of the estimated voting power of the community, representing a notable disparity between voters with the highest and lowest votingpowers. Thisphenomenonisparticularlyevident when examining the mean and median values, which representlessthan0.01%oftheestimatedvotingpower. Thisindicatesthatmorethanhalfofthevoterspossess a voting power that is exceedingly limited, thereby illustrating the existence of a significant inequality in voting power distribution. To further explore this finding, we calculated inequality metrics related to the concentrationofvotingpoweramongvoters. First, we calculated the Gini coefficient, resulting to be 0.9819, indicating that voting power within the DAO is extremely unevenly distributed, with a significant concentration of power among a small number of voters. Subsequently, we conducted an analysis to determine the number of voters needed to hold together more than 50% of the estimated voting power. The results indicated that 15 voters were sufficient to concentrate more than half of the DAO’s voting power, which represents only 0.18% of the total number of voters, illustrating how a relatively rewop gnitov detalumuccA Figure 1. Lorenz curve. Having explored power concentration in a few voters, it is essential to delve deeper into the implicationsofthispowerstructure. Acriticalquestion arises: What is the extent of voter participation among those with greater voting power in governance? Are they the most involved in the decision-making process, or, on the contrary, are they not very participatory? In ordertoascertaintheexistenceofarelationshipbetween participation and power, we calculated the correlation coefficient between the number of times a voter has cast a vote and their median voting power. Pearson’s correlationcoefficientresultedinr=0.0440,confirming nocorrelationbetweenthelevelofparticipationandthe poweravoterhas. The scatter plot in Figure 2 confirms this behavior but also reveals an interesting deviation from it. In particular,intheupperleftpartofthefigure,wecansee thatamongthemostparticipatoryvotersarethosewith verylittlevotingpower.Inaddition,voterswithahigher votingpowerarenotaslikelytoparticipate. Therefore, wecanaffirmthatbeingaverypowerfulvoterdoesnot meanparticipatingthemostingovernanceandthatthey arenotusingtheirpoweronaregularbasistoinfluence theresults. 4.2. Communityconsensus Thenextstepistoexaminetheproposals’outcome in greater detail, focusing on the results obtained to determine whether there is a general consensus around Page 2054 1200 1000 800 600 400 200 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Median voting power of the voter 1e7 retov a yb detsac setov fo rebmuN Figure 2. Ratio between the number of votes cast by a voter and their voting power. anoptionorwhethertherearehighlydisputeddecisions. Firstly, it was observed that 354 proposals (13.9% of the total) were approved unanimously. This indicates thatforasignificantsubsetofdecisions,thecommunity demonstratesahighdegreeofconsensus. 100 80 60 40 20 0 0 20 40 60 80 100 Percentage of votes rewop gnitov fo egatnecreP reinforces the perception of a high degree of cohesion andagreementindecision-making. 4.3. Analysisofvotingtopics The textual data obtained from Decentraland proposalsconstitutesanimportantsourceofinsightinto the dynamics of the community in relation to the key topics, issues, and ideas being discussed. In this study, textual data is analyzed using both Exploratory Data Analytics (EDA) and Topic Modelling to understand whicharethemaintopicsbeingvotedon. The analysis revealed that “interest,” “points,” and “location” were among the four most frequent words. This is due to the fact that Decentraland has a specific category of proposals designated as “Point of Interest” that are always of the form “Add the location X, Y to the Points of Interest” or “Remove the location X, Y from the Points of Interest.” Consequently, we have decided to exclude this category of proposals from our text analysis study, as they introduce a considerable degree of noise. A total of 505 proposals fall within this category (19.83% of the total). Therefore, the text analysiswillencompassatotalof2,042proposals. 4.3.1. ExploratoryDataAnalytics In the present research, EDA techniques were employed to identify top words, which typically refer to the most frequent or commonly occurring words in a dataset. These words are often visualized using techniques such as unigrams, bigrams, trigrams, and wordclouds. The most prevalent words identified by the text analysis are presented in Figure 4. The word “decentraland” is indicative of a strong focus on the platform itself, and “dao” indicates discussions around decentralized governance and identity. It is also Figure 3. Ratio between the percentage of votes known that “dcl“ is an abbreviation for Decentraland. and the voting power received by the winning option. “Wearables” suggests a focus on virtual items within the platform, being natural in a metaverse ecosystem. Thecommunityvotesfairlyevenly,withthewinning Further, “add” suggests a tendency to frequently option receiving an average of 86.73% of the voting consider new features at the platform level. Finally, power used on that proposal and 75.83% of the the top word, “name” could relate to the discussion votes. This indicates that the proposals are not highly of specific names that are used on the platform. contested,anddespitetheunequaldistributionofvoting The remaining top words suggests discussions around power, the outcome does not appear to be imposed. prohibitingoffensivecontent(“ban”),maybeindicative Figure 3 illustrates that the majority of the proposals of a request for financial support to develop projects are situated within the first quadrant, indicating that (“grant”), and highlights the importance of the user the winning option wins both in terms of voting power baseandsocialinteractions(“community”). Lastly,the and votes. This high percentage suggests that even word“linked”mayrefertoconnectionsbetweenvarious in non-unanimous decisions, there is often a clear and elementsinsideoroutsidetheplatform. decisive preference for one option, with the majority While unigrams provide insights into individual of members aligning with the final decision. This word frequencies, analyzing bigrams and trigrams Page 2055 captures word combinations and contextual support mechanisms within the platform to develop relationships, thereby facilitating the identification new projects. Both “linked wearables” and “wearables of common phrases and more profound patterns registry” highlight the significant focus on managing within the text that may not be discernible through and cataloging wearable items, which are crucial the examination of single words. Figure 5 and Figure aspects of the user experience in Decentraland. Lastly, 6 present the top bigrams and trigrams, respectively, the bigram “ban name” suggests banning offensive extractedfromthetextualdataoftheproposals. namesfromtheplatform. Figure 4. Unigram. Figure 6. Trigram. The trigram is illustrated in Figure 6. The results indicate that the proposals under analysis have a strongfocusoneffectivemanagement,governance,and strategicinitiativeswithinthecommunity. Thetrigrams suchas“linkedwearablesregistry,”“wearablesregistry add,” and “wearables curation committee” underline again the platform’s emphasis on the organization and management of wearable items. This implies that a good deal of conversation is on making sure wearables are appropriately categorized and incorporated into the platform. The trigrams “clock decentraland world,” “utc clock decentraland,” and “add utc clock” are Figure 5. Bigram. also noteworthy as they suggest that Decentraland has a large and active community, and may be Thebigramsindicatethatthediscussioniscentered based in different time zones, which might be on the effective management and governance of the important for events and other time-sensitive activities. platformandcommunity,withaparticularfocusonthe Further, key trigrams such as “member revocations expansion of wearable technology. This phenomenon committee”and“confirmedmemberrevocations”point can be observed with the bigrams “Decentraland dao” towardsstructuredproceduresformanagingcommunity and “dao committee”, that highlight the platform’s membership and the enforcement of rules, indicating dedication to decentralized governance by placing an organized approach to governance. Lastly, trigrams a significant focus on governance through the DAO like “dao grants program” and “grant support squad” and committee-based decision-making procedures. highlight the emphasis on financing sources and Furthermore, other key bigrams include the “curation financial assistance that foster platform innovation and committee” and “revocations committee,” which development. In conclusion, the results of trigrams indicatesaneffectiveorganizationalstructurededicated indicate that the Decentraland community is actively tomanagingandfacilitatingthedecisions,andhandling participating in conversations about management, of revocations within the platform through proper governance, and the platform’s strategic growth. The structure, possibly related to governance or financial platform’s ecosystem is managed in an orderly and decisions such as funding a grant. This leads us to systematicmanner,asseenbythenumerousreferences “grants program”, which focuses on the financial to committees and registers, which guarantees the Page 2056 ecosystem’ssustainedgrowthandevolution. greaterdegreeofdispersion,whichmaysuggestthatthe Finally, the results of the word cloud analysis are discussionofthetopicisnotasfocusedontheidentified presented in Figure 7, showing the most occurring topicbyLDA.Thismakessensegiventhenatureofthe words,keytopicsandideasofinteresttothecommunity. topics, which discuss the creation of new projects and The term “decentraland” (and “dcl,” its abbreviation) new content on the platform. Finally, the variety and is one of the most prevalent in the word cloud, specializedfocusareaswithintheclustersemphasizethe indicating that members are engaged in discourse dynamicnatureoftheDecentralandcommunity. pertaining to core ideas and the platform itself. Other terms like “wearables”, “game,” and “nft” are intrinsic Table 2. Topic modeling results. to Decentraland domain and to be expected in a Keywords Topic metaverse. Additionally, “proposals,” “committee,” [‘dcl’,‘community’,‘content’,‘earn’, Community “community,” and “vp” suggest a focus on governance ‘video’,‘part’,‘dapp’,‘library’,‘dao’, andContent and decision-making that is essential for the successful ‘game’] Creation runningoftheplatform. Lastly,“grant,”“new,”“ceate,” [‘voting’,‘dao’,‘change’,‘implement’, DAO and “add” highlight a dynamic and ever-evolving ‘name’,‘ban’,‘add’,‘grant’,‘page’, Governance platform and represent continuous attempts to improve ‘squad’] andVoting andbroadenthevirtualenvironment. [‘decentraland’,‘system’,‘delegate’, Delegation ‘incentives’,‘squad’,‘land’,‘voting’, andIncentives ‘events’,‘add’,‘proposals’] [‘grant’,‘revoke’,‘vp’,‘support’, Grant ‘creators’,‘wearable’,‘request’, Management ‘process’,‘gaming’,‘squad’] andSupport [‘decentraland’,‘decentralands’, ‘squad’,‘protocol’,‘marketplace’, Decentraland ‘new’,‘renewal’,‘studios’, Infrastructure ‘maintenance’,‘builder’] [‘grant’,‘committee’,‘revocations’, Grant ‘resubmission’,‘amount’,‘instead’, Committees ‘proposal’,‘confirmed’,‘category’, andProjects Figure 7. Word Cloud. ‘projects’] [‘wearables’,‘linked’,‘add’,‘dao’, Wearablesand ‘users’,‘district’,‘registry’,‘improve’, UserRegistry 4.4. TopicModeling ‘treasury’,‘vp’] [‘dao’,‘platform’,‘advertising’, Platform ‘engagement’,‘framework’,‘vr’, Development Inaddition,topicmodelingwasemployedtoanalyze ‘team’,‘mobile’,‘strategic’,‘health’] andStrategy thetextdataandextractlatentfeaturesusingtheLatent [‘button’,‘wearables’,‘metaverse’, Metaverse Dirichlet Allocation (LDA) algorithms, generating a ‘development’,‘world’,‘committee’, Development total of 10 topics. Table 2 shows the results of the ‘would’,‘create’,‘meta’,‘curation’] andCuration topic modeling, which indicates that the community [‘decentraland’,‘guidelines’,‘grants’, Guidelines is focused on a wide range of topics, including the ‘mana’,‘program’,‘names’,‘nfts’, andNFT current infrastructure and practices in order to develop ‘nft’,‘make’,‘wearable’] Management andexpandtheplatform. To complement the findings of topic modeling, we conductedaclusteranalysisbasedonLDA,employing thek-meansclusteringalgorithm.Todeterminetheideal numberofclustersfork-means,weutilizedthenumber of topics produced by LDA (Negara et al. (2019)). This permitted the visualization of the topics based on their dispersion using principal component analysis (PCA) as a dimensionality reduction technique. Figure 8 shows that seven of the clusters are clearly defined, despite their proximity to one another. Additionally, the remaining three clusters (i.e. “Decentraland Infrastructure”, “Grant Committees and Projects” and “Metaverse Development and Curation”) exhibit a Figure 8. Cluster analysis. Page 2057

5. Discussion outcomeofvotingprocesses. In conclusion, it can be stated that investigating the The analysis of the distribution of voting power conditionsunderwhichpowerfulvotersusetheirpower withintheDAOrevealsastrongconcentrationofpower, will provide deeper insights into the dynamics of DAO with important implications for the participation of all governanceandhelpidentifypotentialvulnerabilitiesin usersintheactualdecision-makingoftheDAO,farfrom thedecision-makingprocess. a democratic form of governance. The concentration The results of the LDA-topic modeling analysis of voting power in a small group of participants have yielded several key insights for the Decentraland results in a significant imbalance in the distribution of community. The results indicate that the most influenceoverdecision-making,whichiscontrarytothe significant topics are related to the emphasis on principles of democratic governance. This fact raises community and content creation, the governance of fundamentalquestionsabouttherepresentativenessand the DAO, and the voting process. Furthermore, the fairnessofvotingprocessessincefewvoterscouldeasily discourse surrounding the voting and governance imposetheirwill. procedures demonstrates a strong commitment to Despite the potential for these powerful voters decentralized decision-making, ensuring that the to exert considerable influence over the outcome of platform develops in a manner that represents the decisions,ourfindingsindicatethattheydonotregularly collective desires of the users. This commitment to exercisethispower. Nevertheless,althoughtheydonot participatory governance is of the utmost importance currentlyusetheirvotingpowertoinfluenceoutcomes, in order to preserve user involvement, engagement, they retain the ability to do so, thereby having the and confidence. In addition, the findings of topic potentialtounderminecommunitydecisionsatwill. modelinghaverevealedapronouncedfocusonstrategic This dynamic raises a number of important development and infrastructure. Topics such as grant considerations. Firstly, the concentration of power administration, incentives, and delegating suggest does not currently translate into a concentration of that systematic efforts are made to assist creators and influence, as the most powerful voters are not the most encourageactiveengagementontheplatform. participatory. In addition, the high level of consensus However, the analyzed proposals primarily address observed in proposal outcomes may account for the the day-to-day operational aspects of the community limited participation of powerful voters. It is possible andthedevelopmentofnewprojectsthatwillfacilitate that they either agree with the prevailing decisions or the platform’s growth. These proposals lack strategic donotconsiderminordecisionssufficientlyimportantto significance and have no further implications, which engage. Thiscouldalsoindicateadegreeofsatisfaction maybeacontributingfactortothelowparticipationof amongthemostpowerfulvoterswiththestatusquo,or themostpowerfulvoters. arelianceonthebroadercommunitytomakedecisions. Nevertheless, the potential for these influential voters 6. Conclusion to significantly impact outcomes remains, presenting a potentialrisktothedemocraticintegrityoftheDAO. This analysis provides a comprehensive This observation raises an important question for examination of the voting and proposal dynamics futureresearch: underwhatcircumstancesoronwhich within Decentraland DAO. Our findings reinforce specifictopicsdopowerfulvotersleveragetheirposition the ideas of Fritsch et al. (2024) regarding the to impose their criteria in the final outcomes? It is significant concentration of voting power and their crucial to understand when these influential voters are potential capacity to influence decisions. Also, the active,suchaswhethertheyengageonspecifictopicsor nature of distributed governance is sustained by the iftheyvotewiththeintentionofinfluencingtheresults. cryptoanarchist approach that individual freedom is This brings us to the next question. Snapshot a core value to be preserved in the digital sphere, enables anyone to view the outcome of the proposals reducing the capacity of central unique mediators to in real-time. This transparency may allow powerful structure and organize (Chohan (2017)). However, voterstodecidewhethertheirparticipationisnecessary the practical implication of such a principle should go based on their interests, potentially leading to strategic hand-in-handwithalargeparticipationofthemembers use of their power. For instance, they might wait until of communities to ensure a result that represents the the last minute to vote, thereby influencing the results community. But this does not seem to be the case without allowing sufficient time for others to react. in DAOs nor in neighboring uses of cryptoanarchist Furtherresearchisrequiredtoelucidatewhattheimpact philosophy such as Online Creation Communities, of real-time visibility of voting results is on the final where unequal participation is not considered to be a Page 2058 problembutratheranexpectedoutcome(FusterMorell in decision-making. the cases of estonia and (2014)). This phenomenon is similarly observed in the catalonia.Profesionaldelainformacio´n,31(3). study by Pen˜a-Calvin et al. (2024), which analyzed Boyd, D. (2010). Social network sites as networked datafromover10,000DAOsandconcludedthatlarger publics: Affordances, dynamics, and DAOs tend to have greater inequality. Furthermore, implications.InAnetworkedself (pp.47–66). the study found that inequality becomes a common Routledge. characteristicoflargecollaborativeonlinecommunities, Casey,M.J.,&Vigna,P.(2018).Inblockchainwetrust. followingtheironlawofoligarchy. MITTechnologyReview,121(3),10–16. Rather the opposite, the application of distributed Chao, C.-H., Ting, I.-H., Tseng, Y.-J., Wang, B.-W., governancemechanismsinDAOs,opensthedoortothe Wang, S.-H., Wang, Y.-Q., & Chen, M.-C. re-centralization of power in some individuals whose (2022).Thestudyofdecentralizedautonomous condition gets reinforced by the internal dynamics organization (dao) in social network. of governance (Bodo´ et al. (2021)). This situation Proceedings of the 9th Multidisciplinary gives those large-power holders not only a larger and International Social Networks Conference, (on occasions) decisive power but also the possibility 59–65. to strategically use it depending on the situation Chohan, U. (2017). Cryptoanarchism and and the context. This situation sheds doubts on cryptocurrencies.SSRNElectronicJournal. the real capacity of DAOs to fulfill the promise of Cossar, S., Merk, T., Kamalova, J., & De Filippi, becoming a fully democratic environment and makes P. (2024). Proof of humanity: Ethnographic the adoption of this voting method difficult in contexts research of a “democratic” dao. European where inter-individual equality stands as a value over UniversityInstitute. individualfreedom. De Filippi, P., Mannan, M., & Reijers, W. (2020). Blockchain as a confidence machine: The Acknowledgements problem of trust & challenges of governance. TechnologyinSociety,62,101284. The work of Andrea Pen˜a-Calvin was partially Dey, B. L., Yen, D., & Samuel, L. (2020). Digital supported by the project DAOapplications (Spanish consumer culture and digital acculturation. Ministry of Science and Innovation, Grant No.: International Journal of Information PID2021-127956OB-I00), and the Universidad Management,51,102057. Complutense de Madrid - Banco Santander (Grant Egger, R. (2022). Topic modelling: Modelling hidden No.: CT58/21-CT59/21). semantic structures in textual data. In Applied TheworkofDavidDuenas-Cidhasbeenfundedby data science in tourism: Interdisciplinary thePolishNationalResearchCenter(GrantOPUS-20- approaches, methodologies, and applications 2020/39/B/HS5/01661)andEUH2020MSCAProgram (pp.375–403).Springer. (Grantagreementno. 101038055). Ehin, P., Solvak, M., Willemson, J., & Vinkel, P. (2022). Internet voting in estonia 2005–2019: References Evidence from eleven elections. Government InformationQuarterly,39(4),101718. Beckhard, R. (1966). An organization improvement Fan, Y., Zhang, L., Wang, R., & Imran, M. A. (2023). program in a decentralized organization. The Insightintovotingindaos:Conceptualanalysis Journal of Applied Behavioral Science, 2(1), andaproposalforevaluationframework.IEEE 3–25. Network,38(3),92–99. Bodo´, B., Brekke, J. K., & Hoepman, J.-H. (2021). Fritsch, R., Mu¨ller, M., & Wattenhofer, R. (2024). Decentralisation: A multidisciplinary Analyzing voting power in decentralized perspective. Internet Policy Review, 10(2), governance: Who controls daos? Blockchain: 1–21. ResearchandApplications,100208. Borge, R., Balcells, J., & Padro´-Solanet, A. (2023). Fuster Morell, M. (2014). Governance of online Democratic disruption or continuity? creationcommunitiesforthebuildingofdigital analysis of the decidim platform in catalan commons: Viewed through the framework municipalities.AmericanBehavioralScientist, of institutional analysis and developement. 67(7),926–939. Governingknowledgecommons,281. Borge, R., Brugue´, J., & Duenas-Cid, D. (2022). Gil-Garcia,J.R.,Gasco-Hernandez,M.,&Pardo,T.A. Technologyanddemocracy:Thewhoandhow (2020). Beyond transparency, participation, Page 2059 and collaboration? a reflection on the digital society. European Journal of Futures dimensions of open government. Public Research,5,1–13. Performance & Management Review, 43(3), Natale, S., & Ballatore, A. (2014). The web will 483–502. kill them all: New media, digital utopia, Gjøsteen,K.(2016).E-votinginnorway.InReal-world and political struggle in the italian 5-star electronic voting (pp. 119–144). Auerbach movement. Media, Culture & Society, 36(1), Publications. 105–121. Goldberg, M., & Scha¨r, F. (2023). Metaverse Negara, E. S., Triadi, D., & Andryani, R. (2019). governance: An empirical analysis of Topic modelling twitter data with latent voting within decentralized autonomous dirichletallocationmethod.2019International organizations. Journal of Business Research, Conference on Electrical Engineering and 160,113764. ComputerScience(ICECOS),386–390. Halderman, J. A., & Teague, V. (2015). The new south Pen˜a-Calvin, A., Arroyo, J., Schwartz, A., & wales ivote system: Security failures and Hassan, S. (2024). Concentration of power verification flaws in a live online election. and participation in online governance: The E-Voting and Identity: 5th International ecosystem of decentralized autonomous Conference, VoteID 2015, Bern, Switzerland, organizations. Companion Proceedings of the September2-4,2015,Proceedings5,35–53. ACMonWebConference2024,927–930. Han,J.,Lee,J.,&Li,T.(2023).Daogovernance.SSRN Pitt, J., & Diaconescu, A. (2015). Structure and ElectronicJournal. governance of communities for the digital Hassan, S., & De Filippi, P. (2021). Decentralized society. 2015 IEEE International Conference autonomous organization. Internet Policy onAutonomicComputing,279–284. Review,10(2). Sahoo, K., Samal, A. K., Pramanik, J., & Pani, S. K. Jemielniak, D. (2020). Common knowledge? an (2019).Exploratorydataanalysisusingpython. ethnographyofwikipedia.StanfordUniversity InternationalJournalofInnovativeTechnology Press. andExploringEngineering,8(12),4727–4735. Kling, C., Kunegis, J., Hartmann, H., Strohmaier, M., Schneider, N., De Filippi, P., Frey, S., Tan, J. Z., & & Staab, S. (2015). Voting behaviour and Zhang,A.X.(2021).Modularpolitics:Toward power in online democracy: A study of a governance layer for online communities. liquidfeedback in germany’s pirate party. ProceedingsoftheACMonHuman-Computer Proceedings of the International AAAI Interaction,5(CSCW1),1–26. Conference on Web and Social Media, 9(1), Simon,J.,Bass,T.,Boelman,V.,&Mulgan,G.(2017). 208–217. Digital democracy: The tools transforming Licht,N.,Duenas-Cid,D.,Krivonosova,I.,&Krimmer, politicalengagement. R. (2021). To i-vote or not to i-vote: Drivers Vromen, A. (2017). Digital citizenship and political and barriers to the implementation of internet engagement.Springer. voting. Electronic Voting: 6th International Wright, A. (2020). The rise of decentralized Joint Conference, E-Vote-ID 2021, Virtual autonomous organizations: Opportunities Event, October 5–8, 2021, Proceedings 6, andchallenges.Stan.J.BlockchainL.&Pol’y, 91–105. 4,1. Liu,X.(2024).Theillusionofdemocracy—whyvoting Yee, A. (2019). Internet and blockchain technologies: in decentralized autonomous organizations is Authoritarian or democratic? Internet Policy doomedtofail.AvailableatSSRN4441178. Review. Ma, J., Jiang, M., Jiang, J., Luo, X., Hu, Y., Zhou, Y., Zhang, R., Xue, R., & Liu, L. (2019). Security Wang, Q., & Zhang, F. (2024). Demystifying and privacy on blockchain. ACM Computing the dao governance process. arXiv preprint Surveys(CSUR),52(3),1–34. arXiv:2403.11758. Ziegler, C., Miranda, M., Cao, G., Arentoft, G., & Merkle, R. (2016). Daos, democracy and governance. Nam, D. W. (2024). Classifying proposals CryonicsMagazine,37(4),28–40. of decentralized autonomous organizations Minkkinen, M., Auffermann, B., & Heinonen, S. using large language models. arXiv preprint (2017). Framing the future of privacy: arXiv:2401.07059. Citizens’ metaphors for privacy in the coming Page 2060