pith. sign in

arxiv: 2510.05830 · v2 · submitted 2025-10-07 · 💻 cs.CR

Fairness in Token Delegation: Mitigating Voting Power Concentration in DAOs

Johnnatan Messias , Ayae Ide This is my paper

Pith reviewed 2026-05-18 09:34 UTC · model grok-4.3

classification 💻 cs.CR
keywords DAO governancetoken delegationvoting power concentrationinterest alignmentdecentralized autonomous organizationsempirical studyforum analysismisalignment
0
0 comments X

The pith

Delegations in DAOs are frequently misaligned with token holders' priorities and current ranking interfaces worsen voting power concentration.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper studies delegation practices by examining off-chain discussions across 14 DAO forums. It develops a method to connect forum users to their blockchain addresses and applies language models to identify the governance topics they care about from their posts. These extracted interests are then compared to the actual voting records of the delegates those users have chosen. The comparison shows that many delegations do not reflect the priorities expressed in the forums. The authors conclude that ranking-based selection systems amplify this mismatch by favoring a small group of visible delegates regardless of alignment.

Core claim

An empirical analysis of delegation in 14 DAO forums reveals that delegations are frequently misaligned with token holders' expressed priorities and that current ranking-based interfaces exacerbate power concentration. The authors argue that incorporating interest alignment into delegation processes could mitigate these imbalances and improve the representativeness of DAO decision-making.

What carries the argument

A methodology that links forum participants to on-chain addresses, extracts governance interests via large language models from posts, and compares those interests against delegates' historical voting behavior to measure alignment.

Load-bearing premise

Forum posts can be reliably linked to specific on-chain addresses and large language model extractions from those posts accurately capture token holders' true governance priorities.

What would settle it

A large-scale check of linked addresses showing that most delegations match the governance interests extracted from the corresponding forum posts would undermine the misalignment claim.

Figures

Figures reproduced from arXiv: 2510.05830 by Ayae Ide, Johnnatan Messias.

Figure 1
Figure 1. Figure 1: Delegation interface on Tally [29]. The platform displays delegates ranked by default according to total voting power received, as shown in the dropdown menu. This design choice promotes highly visible delegates, such as “Entropy Advisors” and “LobbyFi”, potentially reinforcing a rich-get￾richer dynamic. Token holders are provided with limited contextual information to guide their delegation choices, with … view at source ↗
Figure 2
Figure 2. Figure 2: Overview of our interest-aligned delegation [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Concentration patterns in DAO governance. Figures (a–d) show CDFs of token holders, delegatees, and delegators [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Categorization of proposals in our analyzed forum [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: Prompt used for Proposal Categorization Category Distribution [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 7
Figure 7. Figure 7: , which also guided the models to provide relative impor￾tance and confidence scores as indicators for interpretability, along with the stance on the proposal (positive, negative, or neutral). As shown in [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Dendrogram from hierarchical clustering To inspect the similarities of these clusters, we projected the high-dimensional voter interest vectors into a two-dimensional space using t-SNE, as illustrated in [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: t-SNE plot visualizing cluster assignments of voters. [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Word clouds of top keywords characterizing each [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
read the original abstract

Decentralized Autonomous Organizations (DAOs) aim to enable participatory governance, but in practice face challenges of voter apathy, concentration of voting power, and misaligned delegation. Existing delegation mechanisms often reinforce visibility biases, where a small set of highly ranked delegates accumulate disproportionate influence regardless of their alignment with the broader community. In this paper, we conduct an empirical study of delegation in DAO governance off-chain discussions from 14 DAO forums. We develop a methodology to link forum participants to on-chain addresses, extract governance interests using large language models, and compare these interests against delegates' historical behavior. Our analysis reveals that delegations are frequently misaligned with token holders' expressed priorities and that current ranking-based interfaces exacerbate power concentration. We argue that incorporating interest alignment into delegation processes could mitigate these imbalances and improve the representativeness of DAO decision-making. To support future research, we will release our dataset and code in a public repository.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper conducts an empirical study of delegation practices across 14 DAO forums. It develops a methodology to link forum participants to on-chain addresses, uses LLMs to extract governance interests from posts, and compares these interests against delegates' historical voting records. The central findings are that delegations are frequently misaligned with token holders' expressed priorities and that ranking-based delegation interfaces exacerbate voting power concentration. The authors argue for incorporating interest alignment into delegation mechanisms to improve representativeness and plan to release the dataset and code.

Significance. If the empirical results hold after methodological validation, the work would provide concrete evidence of misalignment in DAO governance and a practical direction for interface and mechanism redesign. This could influence both academic research on decentralized governance and the development of delegation tools in production DAOs, particularly by shifting focus from visibility-based ranking to alignment-based matching.

major comments (1)
  1. [Data collection and address linking section] The linking methodology (described in the section on data collection and address linking) provides no precision, recall, or false-positive metrics and no manual audit of a sample of mappings. Because the misalignment statistics and conclusions about ranking interfaces rest entirely on correctly associating forum posts with specific on-chain addresses, the absence of validation directly undermines the reliability of all downstream LLM-based interest extraction and delegate-behavior comparisons. Even moderate mapping error rates would render the reported misalignment frequencies uninterpretable.
minor comments (2)
  1. [Abstract] The abstract states that the dataset and code will be released but does not provide a repository URL or access instructions; this detail should be added for reproducibility.
  2. [Throughout] Notation for 'delegates,' 'token holders,' and 'forum participants' should be defined once early in the paper and used consistently to avoid ambiguity when discussing the linking step.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading and for highlighting the need for explicit validation of our address-linking procedure. We address this point directly below and commit to strengthening the manuscript accordingly.

read point-by-point responses

  1. Referee: [Data collection and address linking section] The linking methodology (described in the section on data collection and address linking) provides no precision, recall, or false-positive metrics and no manual audit of a sample of mappings. Because the misalignment statistics and conclusions about ranking interfaces rest entirely on correctly associating forum posts with specific on-chain addresses, the absence of validation directly undermines the reliability of all downstream LLM-based interest extraction and delegate-behavior comparisons. Even moderate mapping error rates would render the reported misalignment frequencies uninterpretable.

    Authors: We agree that quantitative validation of the linking step is essential for interpreting the misalignment statistics. The current manuscript describes the heuristic rules and manual spot-checks used to map forum usernames to on-chain addresses but does not report aggregate precision/recall figures or the size of the audited sample. In the revised version we will add a dedicated subsection that (i) details the manual audit protocol, (ii) reports precision, recall, and estimated false-positive rate on a randomly sampled subset of at least 200 mappings, and (iii) discusses the sensitivity of the downstream misalignment frequencies to plausible error rates. This addition will directly address the concern that moderate mapping errors could render the reported statistics uninterpretable. revision: yes

Circularity Check

0 steps flagged

Empirical observational study with no derivations or self-referential reductions

full rationale

The paper conducts an empirical analysis of delegation patterns in 14 DAOs by linking forum participants to on-chain addresses, applying LLM-based interest extraction, and comparing results to delegate voting histories. No equations, fitted parameters, or predictions are defined in terms of the target outputs. No self-citations are invoked as load-bearing uniqueness theorems or ansatzes. The central claims rest on data processing steps whose validity can be assessed externally rather than reducing tautologically to prior quantities by construction. This is a standard empirical workflow with independent content.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on two key untested assumptions about data fidelity and LLM fidelity rather than on free parameters or new invented entities.

axioms (2)
  • domain assumption Forum posts accurately reflect token holders' governance interests
    Invoked when comparing extracted interests against delegate historical behavior.
  • domain assumption Forum participants can be correctly mapped to on-chain addresses
    Required for the empirical comparison between expressed priorities and actual voting.

pith-pipeline@v0.9.0 · 5684 in / 1250 out tokens · 42479 ms · 2026-05-18T09:34:42.135386+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages

  1. [1]

    Hayden Adams, Noah Zinsmeister, Moody Salem, River Keefer, and Dan Robin- son. 2021. Uniswap v3 core. https://web.archive.org/web/20221011013240/https: //uniswap.org/whitepaper-v3.pdf Accessed on July 15, 2025

    work page arXiv 2021

  2. [2]

    Shiri Alouf-Heffetz, Tanmay Inamdar, Pallavi Jain, Nimrod Talmon, and Yash More Hiren. 2024. Controlling Delegations in Liquid Democracy. InProceed- ings of the 23rd International Conference on Autonomous Agents and Multiagent Systems(Auckland, New Zealand)(AAMAS ’24). International Foundation for Autonomous Agents and Multiagent Systems, Richland, SC, 2624–2632

    work page 2024

  3. [3]

    Shiri Alouf-Heffetz, Łukasz Janeczko, Grzegorz Lisowski, and Georgios Papa- sotiropoulos. 2025. The cost perspective of liquid democracy: feasibility and control. InProceedings of the Thirty-Ninth AAAI Conference on Artificial In- telligence and Thirty-Seventh Conference on Innovative Applications of Artifi- cial Intelligence and Fifteenth Symposium on Ed...

    work page doi:10.1609/aaai.v39i13.33478 2025

  4. [4]

    James Austgen, Andrés Fábrega, Sarah Allen, Kushal Babel, Mahimna Kelkar, and Ari Juels. 2023. DAO Decentralization: Voting-Bloc Entropy, Bribery, and Dark DAOs. arXiv:2311.03530 [cs.CR] https://arxiv.org/abs/2311.03530

    work page arXiv 2023

  5. [5]

    Jan Behrens. 2017. The origins of liquid democracy.The Liquid Democracy Journal5, 2 (2017)

    work page 2017

  6. [6]

    Christian Blum and Christina Isabel Zuber. 2016. Liquid democracy: Potentials, problems, and perspectives.Journal of political philosophy24, 2 (2016)

    work page 2016

  7. [7]

    Dion Bongaerts, Thomas Lambert, Daniel Liebau, and Peter Roosenboom. 2025. Vote delegation in defi governance.arXiv preprint arXiv:2503.11940(2025)

    work page arXiv 2025

  8. [8]

    Lewis Carroll. 1884. Principles of Parliamentary Representation. (1884)

    work page

  9. [9]

    Chaotao Chen and Jiangtao Ren. 2017. Forum latent Dirichlet allocation for user interest discovery.Knowledge-Based Systems126 (2017), 1–7

    work page 2017

  10. [10]

    CoinMarketCap. 2025. Top DAO Tokens by Market Capitalization. https:// coinmarketcap.com/view/dao. Accessed July 2025

    work page 2025

  11. [11]

    Abhisek Dash, Abhijnan Chakraborty, Saptarshi Ghosh, Animesh Mukherjee, and Krishna P. Gummadi. 2021. When the Umpire is also a Player: Bias in Private Label Product Recommendations on E-commerce Marketplaces. InProceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency(Virtual Event, Canada)(FAccT ’21). Association for Computing...

    work page doi:10.1145/3442188.3445944 2021

  12. [12]

    Abhisek Dash, Abhijnan Chakraborty, Saptarshi Ghosh, Animesh Mukherjee, and Krishna P. Gummadi. 2022. Alexa, in you, I trust! Fairness and Interpretability Issues in E-commerce Search through Smart Speakers. InProceedings of the ACM Web Conference 2022(Virtual Event, Lyon, France)(WWW ’22). Association for Computing Machinery, New York, NY, USA, 3695–3705...

    work page doi:10.1145/3485447 2022

  13. [13]

    Abhisek Dash, Abhijnan Chakraborty, Saptarshi Ghosh, Animesh Mukherjee, and Krishna P. Gummadi. 2023. FaiRIR: Mitigating Exposure Bias From Related Item Recommendations in Two-Sided Platforms.IEEE Transactions on Computational Social Systems10, 3 (2023), 1301–1313. doi:10.1109/TCSS.2022.3164655

    work page doi:10.1109/tcss.2022.3164655 2023

  14. [14]

    ENS. 2023. Ethereum Name Service (ENS). https://docs.ens.domains Accessed on July 15, 2025

    work page 2023

  15. [15]

    Etherscan. 2025. Etherscan (ETH) Blockchain Explorer. https://etherscan.io. Accessed September 2025

    work page 2025

  16. [16]

    Mengyuan Fang, Jiatian Yu, Chaoyue Gao, and Alvin Leung. 2024. Enhanced Engagement Yet Undermined Decentralization: The Impact of Delegation on User Voting Behavior in DAO. (2024)

    work page 2024

  17. [17]

    Rainer Feichtinger, Robin Fritsch, Lioba Heimbach, Yann Vonlanthen, and Roger Wattenhofer. 2024. SoK: Attacks on DAOs. In6th Conference on Advances in Financial Technologies (AFT 2024) (Leibniz International Proceedings in Informatics (LIPIcs), Vol. 316), Rainer Böhme and Lucianna Kiffer (Eds.). 28:1–28:27. doi:10. 4230/LIPIcs.AFT.2024.28

    work page 2024

  18. [18]

    Emilio Frangella and Lasse Herskind. 2022. Aave v3 technical paper.Visited on 20, 08 (2022), 2022

    work page 2022

  19. [19]

    Bundeszentrale für Politische Bildung. 2025. Wahl-O-Mat. https://www.wahl- o-mat.de/bundestagswahl2025/app/main_app.html. Accessed on February 24, 2025

    work page 2025

  20. [20]

    Matthew Weinberg, and Edward W

    Harry Kalodner, Steven Goldfeder, Xiaoqi Chen, S. Matthew Weinberg, and Edward W. Felten. 2018. Arbitrum: Scalable, private smart contracts. In27th USENIX Security Symposium

    work page 2018

  21. [21]

    Imed Keraghel, Stanislas Morbieu, and Mohamed Nadif. 2024. Beyond Words: A Comparative Analysis of LLM Embeddings for Effective Clustering. InAdvances in Intelligent Data Analysis XXII, Ioanna Miliou, Nico Piatkowski, and Panagiotis Papapetrou (Eds.). Springer Nature Switzerland, Cham, 205–216. 9 J. Messias and A. Ide

    work page 2024

  22. [22]

    Robert Leshner and Geoffrey Hayes. 2019. Compound: The money market protocol

    work page 2019

  23. [23]

    Messari. 2025. Governance Taxonomy — docs.messari.io. https://docs.messari. io/glossary/governance-taxonomy. [Accessed 23-09-2025]

    work page 2025

  24. [24]

    Gummadi, and Patrick Loiseau

    Johnnatan Messias, Vabuk Pahari, Balakrishnan Chandrasekaran, Krishna P. Gummadi, and Patrick Loiseau. 2024. Understanding Blockchain Gover- nance: Analyzing Decentralized Voting to Amend DeFi Smart Contracts. arXiv:2305.17655 [cs.CR] https://arxiv.org/abs/2305.17655

    work page arXiv 2024

  25. [25]

    Kamilla Nazirkhanova, Vrushank Gunjur, X Jesus, and Dan Boneh. 2025. Kite: How to Delegate Voting Power Privately.arXiv preprint arXiv:2501.05626(2025)

    work page arXiv 2025

  26. [26]

    Burak Arda Okutan, Stefan Schmid, and Yvonne-Anne Pignolet. 2025. Democracy for DAOs: An Empirical Study of Decentralized Governance and Dynamic (Case Study Internet Computer SNS Ecosystem).arXiv preprint arXiv:2507.20234(2025)

    work page arXiv 2025

  27. [27]

    Apurva Shah, Axel Abels, Ann Nowé, and Tom Lenaerts. 2025. Artificial Del- egates Resolve Fairness Issues in Perpetual Voting with Partial Turnout. In Proceedings of the ACM Collective Intelligence Conference (CI ’25). Association for Computing Machinery, New York, NY, USA, 71–82. doi:10.1145/3715928.3737482

    work page doi:10.1145/3715928.3737482 2025

  28. [28]

    Tanusree Sharma, Yujin Potter, Kornrapat Pongmala, Henry Wang, Andrew Miller, Dawn Song, and Yang Wang. 2024. Unpacking How Decentralized Au- tonomous Organizations (DAOs) Work in Practice. In2024 IEEE International Conference on Blockchain and Cryptocurrency (ICBC)

    work page 2024

  29. [29]

    Tally. 2025. Arbitrum Delegates. https://www.tally.xyz/gov/arbitrum/delegates. Accessed on February 24, 2025

    work page 2025

  30. [30]

    Tally. 2025. Compound Delegates. https://www.tally.xyz/gov/compound/ delegates. Accessed on February 24, 2025

    work page 2025

  31. [31]

    Tally. 2025. ENS Delegates. https://www.tally.xyz/gov/ens/delegates. Accessed on February 24, 2025

    work page 2025

  32. [32]

    Tally. 2025. Tally: Institutional-Grade Tokens. https://www.tally.xyz. Accessed on July 17, 2025

    work page 2025

  33. [33]

    Tally. 2025. Uniswap Delegates. https://www.tally.xyz/gov/uniswap/delegates. Accessed on February 24, 2025

    work page 2025

  34. [34]

    Tally. 2025. ZKsync Delegates. https://www.tally.xyz/gov/zksync/delegates. Accessed on February 24, 2025

    work page 2025

  35. [35]

    Eric Tang, Bangding Yang, and Xingyou Song. 2025. Understanding LLM Em- beddings for Regression.Transactions on Machine Learning Research(2025). https://openreview.net/forum?id=Wt6Iz5XNIO

    work page 2025

  36. [36]

    The web3.py team. 2025. Web3.py documentation. https://web3py.readthedocs.io. Accessed September 2025

    work page 2025

  37. [37]

    Lukas Weidener, Fabio Laredo, Kishore Kumar, and Karlin Compton. 2025. Del- egated voting in decentralized autonomous organizations: a scoping review. Frontiers in Blockchain8 (2025), 1598283

    work page 2025

  38. [38]

    Inc Wikimedia Foundation. 2025. The rich get richer and the poor get poorer. https://en.wikipedia.org/wiki/The_rich_get_richer_and_the_poor_get_ poorer. Accessed on February 24, 2025. A Delegation Graph Statistics Table 6 reports detailed structural statistics for delegation graphs across the five studied protocols. Each graph models addresses as nodes an...

    work page 2025