Research Summary and AMA with Kelsie Nabben (RMIT University): Can Algorithmic Governance turn a DAO into a Panopticon?

Kelsie Nabben is a Researcher at RMIT University’s Blockchain Innovation Hub, a scholarship recipient from the Centre for Automated Decision Making & Society and a member of the Digital Ethnography Research Centre.

She recently sat down with Chainlink Labs to discuss her paper on algorithmic governance in DAOs entitled “Is a DAO a Panopticon? Algorithmic governance as creating and mitigating vulnerabilities in ‘Decentralised Autonomous Organisations’”.

She has offered to answer questions about her paper here for a limited time so please don’t hesitate to ask any questions you may have!

Description

This interview is a recent episode from the Chainlink Research Report series which features short presentations of working papers by blockchain scholars in computer science, economics, and related fields.

In this episode, RMIT University researcher Kelsie Nabben discusses her latest paper exploring some of the challenges and risks associated with algorithmic governance in DAOs.

Video:

Take-aways:

  • Algorithmic governance, which frequently uses machine learning, solves many security issues inherent in DAOs.
  • At the same time, algorithmic governance creates monitoring systems that change DAOs in fundamental ways, increasing the risk that they will function as panopticons.
  • To strike a balance between security and privacy with algorithmic governance in DAOs, it is essential that the DAO community participate in shaping DAO governance.

Kelsie Nabben background:

Kelsie Nabben is a Researcher at RMIT University in Melbourne, Australia. She conducts ethnographic research on decentralized technology communities, digital infrastructure, blockchain community culture, and algorithmic governance. Kelsie is a recipient of a PhD scholarship at the RMIT University Centre of Excellence for Automated Decision-Making & Society, and a researcher in the Digital Ethnography Research Centre and Blockchain Innovation Hub.

Some of Kelsie’s work:

Papers

Substack - “on the cataclysmia of digital infrastructure

Relevant links and contact info for Kelsie:

5 Likes

I sincerely love this framing and can’t wait to listen to this interview! Applying the panopticon framework seems appropriate to highlight the potential pitfalls of this type of transparency, and I think the framing is succinct and clear. Additionally, this seems to be the least confrontational means of raising potential problems with DAOS.

3 Likes

@kelsienabben - This was a fantastic presentation, and @jasonanastas thank you so much for the interesting questions and excellent hosting. K.,I really liked your analysis of how token ownership can garble the participatory relationship between labor and ownership. It certainly reflects what I’ve encountered in these projects. Do you think DAOs then falling into the same structure as any other capitalist(?) organization or do DAOs as human-machine chimeras present a new paradigm? Has anyone made a connection with Haraway’s Cyborg Manifesto?

I’m also curious whether you’ve encountered a difference in a community being observed by humans versus those observed by algorithms (i.e. are electronic panopticons scarier than straw bosses?). When you were looking at some of the monitoring and incentivization systems like SourceCred and Coordinape, did you notice significant shifts in people’s behavior once these systems were implemented? Were there algorithmic systems of demerits (perhaps similar to ones in classroom tracking apps)? To your points about maintaining digital infrastructure - did you see any difference in the way ‘management’ was perceived in these organizations versus someone doing the digital equivalent of digging ditches?

Thank you so much for taking the time to answer some questions!

2 Likes

Hi James,
Thank you so much for engaging with this piece. I’ve spent a couple of days sitting with the questions you’ve asked, mostly because I would most like to describe and link to other works! (I will share links in this thread as possible, noting that this is a working paper, and there is ongoing research in the Metagov community with projects like Sourcecred).

The piece mostly points out the panopticon dynamic of DAOs in that “trustless”, scalable coordination depends on governance by algorithms, which places people back under a political system, when they sought autonomy (as in, self-governance). I have tried to constructively point out these dynamics in DAOs, to as to help designers reflect on the algorithmic rules they decide on. One of the biggest surprises of this research has been the lack of awareness to, or conversation on, this surveillance by algorithms. The general consensus seems to be acceptance that they are objective, not subjective, and therefore can’t be questioned or changed.

In terms of what’s scarier, that depends very much on your imagination of Artificial General Intelligence and what it means for this to be decentralised. I tackle these issues in the two papers that follow (under review at present).

3 Likes

Kelsie, thank you so much for that fantastic presentation!

To give some context, I came up with “Decentralized Conglomerate Theory” as a counterpoint to a DAO with the understanding that DAOs really don’t have a good mechanism for keeping institutional memory.

With that said, I have seen many DAO frameworks and theoretical constructions that all seem to assume a base level of understanding for the user. My question is: considering DAOs were conceptualized by people that had never led a small group (let alone a global organization) to understand group dynamics; were DAOs effectively just reproducing the same governance system we have and replacing a central authority with an algorithm?

As far as I can tell, DAOs do not actually seem to have innovated in “governance” and have just adopted the pre-existing governance model. Further, the notion that an algorithm removes a centralizing vector assumes the algorithm is “unbiased”. While the notion that “anyone can contribute to an open-source code base” is philosophically true, in practice the contributors to blockchain code bases are a very small part of the community.

Also, as a cybersecurity specialist I know that we are taught that “physical security” is equally if not more important than digital security. The notion that AI can solve problems is a concept that is NOT coming from “cybersecurity specialists” and is one coming from “cypher-punks” that aren’t actually cybersecurity specialists. Are you seeing social engineering being referenced more in your DAO research, or is that still being ignored by non-cybersecurity specialists? I would argue that the social engineering attack vector is likely the most dangerous and the most likely vector for attack in a DAO setting, i.e. Discord phishing, email phishing, fake websites attempting to gain access to web3 wallet, etc…

In other words: If cybersecurity specialists are saying that social engineering is a problem, why are discussions of DAOs and their security seemingly devoid of the discussion of the physical access point or social engineering attacks?

Access control is an integral part of cybersecurity. It is hardly discussed in the crypto space. My final question is, how are we supposed to seriously discuss cybersecurity if the people leading the conversation (Larimer, Buterin, etc…) have no understanding of social engineering attacks or how to protect physical access points?

In other words, without proper access control education, could a DAO ever be feasible since humans as the weakest link will likely be the attack vector?

Further: Is a DAO more likely to be a panopticon when users have not been educated enough to recognize a panopticon potentially due to the presence of propagandistic social-engineering keeping them uninformed (or more appropriately “obedient good citizens”)? Is it that the DAO is the panopticon or that this particular society cannot form governance mechanisms without surveillance being the default due to the habituated nature of the nationally constructed surveillance states?

3 Likes

@kelsienabben, thank you for your fascinating paper and the video with @jasonanastas of ChainLink.

@jmcgirk has raised this question: Is it an engineering problem or a social problem? The prominence of “algorithms” makes me think that we’re dealing with it as an engineering problem—at least here and for now. (Jeremy Bentham probably felt it was a social problem himself, but he applied plenty of engineering to “solving” it.)

“Reputation” derives from personal history, yet if you believe in “education” you accept that people will change their minds—sometimes about significant things. At the same time, human beings (and probably many animals) revere “precedent” and past-performance, and often view personal growth and development as irrelevant or illusory.

For this reason we turn to something “outside ourselves,” namely design and engineering, ideally with serious math behind it. The most revered current example is probably the domain of zero-knowledge proofs, which allow someone to “prove” something—say, their identity—without revealing what that thing actually is.

What do you and Jason think about deploying something like “moon-math based” tools to maintain security with anonymity in DAOs? Is it just more technologist self-delusion, or might it point to a workable solution?

3 Likes

Hi @Larry_Bates.
I’m really grateful for your perspective here. To respond to a few key points made:

  1. I agree - DAOs are discovering governance. The economists at RMIT Blockchain Innovation Hub that research team I’m in does great work on the institutional economics of blockchains and DAOs, and has for a number of years. I find that drawing on existing institutional dynamics to explain and compare certain aspects of DAOs is useful.

  2. “Are you seeing social engineering being referenced more in your DAO research, or is that still being ignored by non-cybersecurity specialists?”. I’d be very interested to hear more on your perspective on this, including AI security. I argue that the purpose of securing digital infrastructure is people before protocol in this paper on “Blockchain Security as People Security”.

  3. You are correct. The society makes the DAO. A DAO, or any institution, is not inherently a panopticon. In fact, as an observation, some DAOs have very relationally based accountability structures in working groups (e.g. Commons Stack). They draw on Ostrom’s “Principles of the Commons” for multipolarity and scalability here.

2 Likes

Hi Ralph,
Thank you for your considered comments and questions.

First of all, I’m not trying to diagnose a problem in this piece, I’m trying to point to what generations of technologists that understand information technology tools in the context of social political systems have said; ‘be careful what you build and how you build it’.

In saying this, all engineering starts with subjective norms, beliefs, and values. In the discussion section of the piece, I point to the idea of “algorithms as policy”, in that writing algorithms in code is like writing laws, they are not “outside ourselves” or our societies. This perspective is a reflection on the blockchain mentality that society can be “engineered” because algorithms, and maths, is objective.

In saying that, the task at hand is to engineer social systems using technology, but being very clear about why, and what principles guide this perspective. For example, ZKPs are great if you want privacy-preserving identification systems, but would have very different social implications if they are used to determine certain parameters (e.g. age verification for access), or identity completely (e.g. Stewart Brand and Howard Reingold vehemently warn against absolute anonymity, versus pseudonymity, this based on their experience of early online communities)

3 Likes

@kelsienabben, thank you for the illuminating response.

3 Likes

Hi @kelsienabben ,
thank you for such a fascinating presentation and a paper full of exceptional points! I really like your points that arouse the awareness of the influence of cultures in DAOs, which give further DAOs experiments progress and research a crucial sign and guide.

Your observation of some DAOs can question and re-evaluate their algorithmic policies correspond to the point <An Exploration of Governing via IT in Decentralized Autonomous Organizations> (2021) proposed, that DAOs have tried to build more pluralistic and decentralized forms of algorithmic management through the mechanism of “taming algorithmic power.” When you are looking at some DAO cases that allow members to shape their rules, did you observe the boundary of this rule re-produce? Are there boundaries to the re-oriented of rules in these DAO communities’ cultures explicitly or impliedly?

Thanks for @rlombreglia’s inspiration. I also think the tension between the code in DAOs and human power has some similarities to a hybrid of civil law influence and common law influence. The civil law system usually implements well-framed laws and creeds with statutory law, while the common law system usually reveres precedents, which are shaped continually by the discretion of judges. Actually, I regard the re-oriented of rules in DAOs more like the operation of the amendment process in civil law, since there have a set of rules already been set by the creators, and then the members discuss and affect its shape like parliaments, which is influenced significantly by the culture and interest behind it as well. Did you notice some similarities between the rules re-oriented process in DAOs and the traditional law amendment process?

Thank you so much for taking the time to answer questions!

Reference
Mini, Tobias; Ellinger, Eleunthia Wong; Gregory, Robert W.; and Widjaja, Thomas, “An Exploration of Governing via IT in Decentralized Autonomous Organizations” (2021). ICIS 2021 Proceedings. 1.
https://aisel.aisnet.org/icis2021/gen_topics/gen_topics/1

1 Like

Thanks Astrid.

Culture is a difficult one. It’s very intangible, yet, social factors are highly influential in communities.

I tried to read the link you shared but its pay walled. Thus, I’m unsure what is meant by “taming algorithmic power” in this paper, however, your questions that follow are very relevant. The next paper I’ve drafted is on if and how DAOs can collectively shape algorithmic rules.

Regarding DAOs and law, there’s some great scholarship in this area (see Di Fillipi and Hassan, for example). From an ethnographic perspective, a repeated theme in DAO communities is an unwillingness to question algorithmic rules, harking back to the idea that “code is law”, rather than a subjective set of choices. I write on this with Dr Michael Zargham here: Algorithms as Policy: | Kelsie Nabben.

3 Likes

Hello, how can I get access to these materials?
I am interested in the topic written about in the text

1 Like

Hey @GloriaOkoba, here’s a link to the paper discussed in the post: Is a DAO a Panopticon? Algorithmic governance as creating and mitigating vulnerabilities in 'Decentralised Autonomous Organisations' by Kelsie Nabben :: SSRN.

You also might want to subscribe to @kelsienabben Substack here: https://kelsienabben.substack.com/. At the top of the post there are also some other resources related to some of the work that Kelsie is doing.

Hope that helps.

2 Likes

This is a great question Ralph. I think this is possible using methods that are currently used in machine learning and social science for data analysis under privacy constraints. I’m not exactly sure how these methods would be employed but this is where I would start.

Here are a few resources discussing these methods:

3 Likes

Hi. I’m not completely sure what you mean by “the materials,” but Kelsie Nabben’s “Panopticon” paper is linked to at the beginning of this summary.

Some other scholarly papers by her are here: Author Page for Kelsie Nabben :: SSRN

That should give you a start. Does this answer your question?

1 Like

Thanks , this will be extremely helpful

Oh yeah, I just didn’t catch a glimpse of her very own Article. I read a couple of comments and it became eye catching to me. Thanks for your guidance .

1 Like

Thanks for this I learnt so much but I have a question.

decentralized autonomous organisations are essentially a corporate governance structure built around crypto , this is the idea I was exposed to during my research .
But to get access to this organisation one must have enough tokens to buy their way in, my concern is this the ideology of cryptocurrency hasn’t been embraced in a lot of countries yet, I’m from Nigeria an African country and in my continent we only have three countries that have central banks for digital currencies

I don’t know what DAO plans to attain in the future with Africa but wouldn’t this discourage a lot of Africans if they find out they are exposed to Sybil attacks?

wouldn’t that bring about a reduction of value for some cryptocurrencies because most Africans might most likely sell their coins out of panic?

Please :pray: this question calls for concern, I strongly believe it does.

@jasonanastas Thanks for your kind response and pointers to the papers. Very interesting.

3 Likes

@kelsienabben , thanks for your response and sharing the enlightening article. I’m very much looking forward to reading your next paper. I’m also curious what you’re referring to the “unsafety” when you mentioned “Related topics which we intend to explore include: how do we ‘safely’ test algorithmic policies” in that article?

3 Likes