Subscribe to taotechaching
Subscribe to taotechaching
A brief note on the collective exocortex to come
From a 2019 IEEE paper titled Decentralized Construction of Knowledge Graphs for Deep Recommender Systems Based on Blockchain-Powered Smart Contracts The “collective exocortex” is a term I’m using as a placeholder for an idea that I predict will begin with the benign interlinking of individuals
Digital land is not land
A brief note on the collective exocortex to come
From a 2019 IEEE paper titled Decentralized Construction of Knowledge Graphs for Deep Recommender Systems Based on Blockchain-Powered Smart Contracts The “collective exocortex” is a term I’m using as a placeholder for an idea that I predict will begin with the benign interlinking of individuals
Digital land is not land
Share Dialog
Share Dialog
<100 subscribers
<100 subscribers
From Public Choice — A Primer by Eamonn Butler
Voting Paradoxes in Action
Indeed, as Condorcet pointed out, some systems could produce almost any result. Rock might lose to Paper, and Paper to Scissors, but Scissors would still be defeated by Rock. The outcome depends on how the election is managed. If there is someone who can set the order in which the votes are taken – say, the chair of a committee that has to choose between several options – that agenda-setter can rig the order in which the votes are taken in order to ensure that his or her own preferences prevail, regardless of what other people want. As Duncan Black calculated – and recent analysis confirms – the more options that are on offer, and the more electors there are voting, the more severe this cycling paradox becomes, making it of particular importance in the large and complex political systems of today.
From Wikipedia on Agenda-setting theory:
Agenda-setting describes the "ability (of the news media) to influence the importance placed on the topics of the public agenda". The study of agenda-setting describes the way media attempts to influence viewers, and establish a hierarchy of news prevalence.
The theory also suggests that media has a great influence to their audience by instilling what they should think about, instead of what they actually think. That is, if a news item is covered frequently and prominently, the audience will regard the issue as more important.
Most researches on agenda-setting are based on the following:
the press and the media do not reflect reality; they filter and shape it;
media concentration on a few issues and subjects leads the public to perceive those issues as more important than other issues.
The relatively recent emergence of several cryptographic primitives can enable decentralized platforms for revealing popular opinion on a global scale. Existing social media-based methods for establishing public consensus are flawed for multiple reasons: the moderation practices of social media platforms are often arbitrary and opaque, there exists path dependence in what topics achieve virality, and the viral nature of social media platforms tend to amplify perspectives from high follower accounts rather than optimizing for exposure to a diversified, pluralistic set of viewpoints, etc. Furthermore, the centralization of these platforms poses questions about their credible neutrality and trustworthiness, thereby restricting the range of perspectives that can be explored de facto (especially salient with respect to those viewpoints that are antithetical to the platforms themselves).
In other words, the agenda-setting process has always had and continues to have the effect of advantaging special interests over the general public because of the inherent structural advantages enjoyed by certain people along the media value chain. One potential solution towards remedying the systematic distortion of public opinion might be the adoption of decentralized e-voting protocols that employ digital identifiers, verifiable randomness, and privacy preservation techniques:
digital identifiers to prove that a prospective voter/proposer meets certain criteria desired by the protocol (i.e., is a unique human that is not voting/proposing using more than one account); the compositing of different identifiers can permit for preservation of identity and unnecessary information (through zero-knowledge schemes) while also providing certain guarantees for unique personhood (e.g., Gitcoin Grants composites various identifiers together to increase contribution matching in their quadratic funding rounds)
verifiable randomness to be able to allocate subsets of the overall set of crowdsourced proposals to voters over the iterative voting process
privacy preservation for both identity and for ensuring secret ballots; the preservation of privacy through zero-knowledge proofs may allow for people to vote their true preferences and deter vote buying [i.e., Potential vote buyers will be disincentivized from purchasing votes if there’s no way to verify the seller took their desired actions; Vitalik talks about this concept in several places, but I can’t find them right now]
And of course, all of the code governing the protocol would be open source [and should be relatively simple] and would ideally be hosted on decentralized frontends [see: Skynet blog post].
But enough level-setting, here’s the idea in bullet-point form:
there is a decentralized voting protocol that prompts members at the beginning of every calendar month to submit and vote on what they think the most important issue [the prompt choice is variable — it could be another prompt like “What was your favorite thing you’ve read or listened to this month?”] for the community/company/city/country/world/etc. is currently; explicit agenda-setting, in other words
10,000,000 [or another arbitrarily large number up to 8 billion] participants in a decentralized voting protocol, all verified by some combination of PoH/SBTs/POAPs/etc. that all other participants find acceptable [i.e., you can create a decentralized voting protocol that accepts some centralized, opaque identifier as a 1-to-1 equivalent for a more credible identifier, but people will be disincentivized to join such a protocol under such conditions]
there is an initial stage that lasts 7 days [or 1 day, whatever amount of time that is sufficient to ensure that every one gets a chance to submit their proposal to the protocol, e.g. if members are distributed geographically around the world] in which each of the 10,000,000 unique individuals must provide a submission
the second stage involves each participant receiving 5 submissions [could as little as 2 or significantly more than 5, but it’s difficult for people to preference rank a set of 50 things — 4 to 7 is probably the ideal number] from the other 9,999,999 individuals and ranking the 5 submissions from 1 to 5 according to their personal preferences — if this step is done once, then each of the 10,000,000 submissions is guaranteed to be independently assessed 5 times [that is, assuming every participant engages in voting]; this can be repeated such that each submission is independently assessed (N x 5) times, where N is the amount of times each participant has to rank a set of 5 submissions; if N = 5, then each proposal is independently assessed and ranked 25 times, thereby making the assessment of each proposal more reliable; this ranking/voting period could be a day or two [or whatever] to give people a chance to submit their response to the protocol
those who fail to complete the ranked choice voting stage of other proposals have their submission automatically disqualified for the next round of voting; in order to disincentivize people from failing to rank proposals, different voting protocols can implement staking/slashing mechanisms and/or ban those identities and wallets that repeatedly fail to vote [
The last bullet point illustrates how this iterative ranking scheme could be used to elicit and incorporate input from the entirety of humanity [that is, in a hypothetical world where we have full internet penetration, phone/computer penetration, and have onboarded literally every body on Earth onto some trusted digital identification scheme that can’t be abused or Sybil attacked] while being entirely manageable on the individual level [it’d be tantamount to answering a few multiple choice response questions every day for a couple of days]. Imagine if the entire world could have a conversation with itself. If humanity had a mirror absent of distortions with which it could look upon itself with. More likely, however, such a scheme would be implemented first on a country-level (if ever) by countries that have high internet/smartphone adoption, an ideological affinity for digital democracy, and a respect for citizen privacy [so basically just Taiwan and Estonia, for now].
That’s pretty much the meat of what I wanted to communicate with this post. Everything that follows from here touches on some of the finer (see: pedantic) points around this idea. In no particular order:
the scoring scheme can be dynamic and could be designed to incorporate factors like the concept of a “correlation score” (from Weyl et al.’s Decentralized Society: Finding Web3’s Soul) in which responses from accounts exhibiting high correlation in their history of responses/votes are dscounted in order to optimize towards a plurality of views; such a scoring scheme would introduce an especially complex layer to the protocol and would require participants to understand the concept in order for it to have broad-based legitimacy
given that most of the protocol submissions will have indepedently sourced submissions that are similar (or even identical) to them, each voting/ranking stage can be changed to incorporate the ability for voters/rankers to identify similar prompts and merge or cluster them together; we can imagine people don’t have to rank the text of single entries but end up ranking clusters of entires that exist within the same semantic and conceptual space [i.e., “I’m really concerned about XYZ” is grouped with “XYZ is the most concerning thing to me this month”]
the prompt can be more dynamic than what I’m suggested; it may be that the person who provided the winning submission to the previous prompt can set the following prompt [e.g., They get to ask a question to the voting protocol]
bootstrapping a large voting protocol may involve activation upon reaching a pre-set minimum threshold of members [e.g., begin the first vote only after 1,000,000 people have joined the protocol]
that the profiles of those who submit the responses are hidden isn’t a necessary condition for a voting protocol; it may be that submissions can be attached to an arbitrary set of information about the submitter; there could be designs that allow the submitter of the winning response(s) the option (but not obligation) to to publicly verify that they were the ones who submitted it
the literature around iterative voting talk about concepts like what “synchronous” means in the context of voting, convergence, and acyclicality; all of these concepts are important for understanding the point of iterative voting but I will write a more systematic review of these concepts in the context of my idea when I personally understand them better
Additional resources:
A Scalable Implementation of Anonymous Voting over Ethereum Blockchain (2021) by Song et al.
Wikipedia: Agenda-setting theory
Iterative Local Voting for Collective Decision-making in Continuous Spaces (2019) by Garg et a.l
Analysis of Equilibria in Iterative Voting Scheme (2015) by Rabinovich et al.
Acyclic Games and Iterative Voting (2016) by Meir et al.
Iterated Majority Voting (2009) by Airiau et al.
Cycles in synchronous iterative voting: general robustness and examples in Approval Voting (2022) by Kloeckner
Decentralized Society: Finding Web3’s Soul (2022) by Weyl et al.
From Public Choice — A Primer by Eamonn Butler
Voting Paradoxes in Action
Indeed, as Condorcet pointed out, some systems could produce almost any result. Rock might lose to Paper, and Paper to Scissors, but Scissors would still be defeated by Rock. The outcome depends on how the election is managed. If there is someone who can set the order in which the votes are taken – say, the chair of a committee that has to choose between several options – that agenda-setter can rig the order in which the votes are taken in order to ensure that his or her own preferences prevail, regardless of what other people want. As Duncan Black calculated – and recent analysis confirms – the more options that are on offer, and the more electors there are voting, the more severe this cycling paradox becomes, making it of particular importance in the large and complex political systems of today.
From Wikipedia on Agenda-setting theory:
Agenda-setting describes the "ability (of the news media) to influence the importance placed on the topics of the public agenda". The study of agenda-setting describes the way media attempts to influence viewers, and establish a hierarchy of news prevalence.
The theory also suggests that media has a great influence to their audience by instilling what they should think about, instead of what they actually think. That is, if a news item is covered frequently and prominently, the audience will regard the issue as more important.
Most researches on agenda-setting are based on the following:
the press and the media do not reflect reality; they filter and shape it;
media concentration on a few issues and subjects leads the public to perceive those issues as more important than other issues.
The relatively recent emergence of several cryptographic primitives can enable decentralized platforms for revealing popular opinion on a global scale. Existing social media-based methods for establishing public consensus are flawed for multiple reasons: the moderation practices of social media platforms are often arbitrary and opaque, there exists path dependence in what topics achieve virality, and the viral nature of social media platforms tend to amplify perspectives from high follower accounts rather than optimizing for exposure to a diversified, pluralistic set of viewpoints, etc. Furthermore, the centralization of these platforms poses questions about their credible neutrality and trustworthiness, thereby restricting the range of perspectives that can be explored de facto (especially salient with respect to those viewpoints that are antithetical to the platforms themselves).
In other words, the agenda-setting process has always had and continues to have the effect of advantaging special interests over the general public because of the inherent structural advantages enjoyed by certain people along the media value chain. One potential solution towards remedying the systematic distortion of public opinion might be the adoption of decentralized e-voting protocols that employ digital identifiers, verifiable randomness, and privacy preservation techniques:
digital identifiers to prove that a prospective voter/proposer meets certain criteria desired by the protocol (i.e., is a unique human that is not voting/proposing using more than one account); the compositing of different identifiers can permit for preservation of identity and unnecessary information (through zero-knowledge schemes) while also providing certain guarantees for unique personhood (e.g., Gitcoin Grants composites various identifiers together to increase contribution matching in their quadratic funding rounds)
verifiable randomness to be able to allocate subsets of the overall set of crowdsourced proposals to voters over the iterative voting process
privacy preservation for both identity and for ensuring secret ballots; the preservation of privacy through zero-knowledge proofs may allow for people to vote their true preferences and deter vote buying [i.e., Potential vote buyers will be disincentivized from purchasing votes if there’s no way to verify the seller took their desired actions; Vitalik talks about this concept in several places, but I can’t find them right now]
And of course, all of the code governing the protocol would be open source [and should be relatively simple] and would ideally be hosted on decentralized frontends [see: Skynet blog post].
But enough level-setting, here’s the idea in bullet-point form:
there is a decentralized voting protocol that prompts members at the beginning of every calendar month to submit and vote on what they think the most important issue [the prompt choice is variable — it could be another prompt like “What was your favorite thing you’ve read or listened to this month?”] for the community/company/city/country/world/etc. is currently; explicit agenda-setting, in other words
10,000,000 [or another arbitrarily large number up to 8 billion] participants in a decentralized voting protocol, all verified by some combination of PoH/SBTs/POAPs/etc. that all other participants find acceptable [i.e., you can create a decentralized voting protocol that accepts some centralized, opaque identifier as a 1-to-1 equivalent for a more credible identifier, but people will be disincentivized to join such a protocol under such conditions]
there is an initial stage that lasts 7 days [or 1 day, whatever amount of time that is sufficient to ensure that every one gets a chance to submit their proposal to the protocol, e.g. if members are distributed geographically around the world] in which each of the 10,000,000 unique individuals must provide a submission
the second stage involves each participant receiving 5 submissions [could as little as 2 or significantly more than 5, but it’s difficult for people to preference rank a set of 50 things — 4 to 7 is probably the ideal number] from the other 9,999,999 individuals and ranking the 5 submissions from 1 to 5 according to their personal preferences — if this step is done once, then each of the 10,000,000 submissions is guaranteed to be independently assessed 5 times [that is, assuming every participant engages in voting]; this can be repeated such that each submission is independently assessed (N x 5) times, where N is the amount of times each participant has to rank a set of 5 submissions; if N = 5, then each proposal is independently assessed and ranked 25 times, thereby making the assessment of each proposal more reliable; this ranking/voting period could be a day or two [or whatever] to give people a chance to submit their response to the protocol
those who fail to complete the ranked choice voting stage of other proposals have their submission automatically disqualified for the next round of voting; in order to disincentivize people from failing to rank proposals, different voting protocols can implement staking/slashing mechanisms and/or ban those identities and wallets that repeatedly fail to vote [
The last bullet point illustrates how this iterative ranking scheme could be used to elicit and incorporate input from the entirety of humanity [that is, in a hypothetical world where we have full internet penetration, phone/computer penetration, and have onboarded literally every body on Earth onto some trusted digital identification scheme that can’t be abused or Sybil attacked] while being entirely manageable on the individual level [it’d be tantamount to answering a few multiple choice response questions every day for a couple of days]. Imagine if the entire world could have a conversation with itself. If humanity had a mirror absent of distortions with which it could look upon itself with. More likely, however, such a scheme would be implemented first on a country-level (if ever) by countries that have high internet/smartphone adoption, an ideological affinity for digital democracy, and a respect for citizen privacy [so basically just Taiwan and Estonia, for now].
That’s pretty much the meat of what I wanted to communicate with this post. Everything that follows from here touches on some of the finer (see: pedantic) points around this idea. In no particular order:
the scoring scheme can be dynamic and could be designed to incorporate factors like the concept of a “correlation score” (from Weyl et al.’s Decentralized Society: Finding Web3’s Soul) in which responses from accounts exhibiting high correlation in their history of responses/votes are dscounted in order to optimize towards a plurality of views; such a scoring scheme would introduce an especially complex layer to the protocol and would require participants to understand the concept in order for it to have broad-based legitimacy
given that most of the protocol submissions will have indepedently sourced submissions that are similar (or even identical) to them, each voting/ranking stage can be changed to incorporate the ability for voters/rankers to identify similar prompts and merge or cluster them together; we can imagine people don’t have to rank the text of single entries but end up ranking clusters of entires that exist within the same semantic and conceptual space [i.e., “I’m really concerned about XYZ” is grouped with “XYZ is the most concerning thing to me this month”]
the prompt can be more dynamic than what I’m suggested; it may be that the person who provided the winning submission to the previous prompt can set the following prompt [e.g., They get to ask a question to the voting protocol]
bootstrapping a large voting protocol may involve activation upon reaching a pre-set minimum threshold of members [e.g., begin the first vote only after 1,000,000 people have joined the protocol]
that the profiles of those who submit the responses are hidden isn’t a necessary condition for a voting protocol; it may be that submissions can be attached to an arbitrary set of information about the submitter; there could be designs that allow the submitter of the winning response(s) the option (but not obligation) to to publicly verify that they were the ones who submitted it
the literature around iterative voting talk about concepts like what “synchronous” means in the context of voting, convergence, and acyclicality; all of these concepts are important for understanding the point of iterative voting but I will write a more systematic review of these concepts in the context of my idea when I personally understand them better
Additional resources:
A Scalable Implementation of Anonymous Voting over Ethereum Blockchain (2021) by Song et al.
Wikipedia: Agenda-setting theory
Iterative Local Voting for Collective Decision-making in Continuous Spaces (2019) by Garg et a.l
Analysis of Equilibria in Iterative Voting Scheme (2015) by Rabinovich et al.
Acyclic Games and Iterative Voting (2016) by Meir et al.
Iterated Majority Voting (2009) by Airiau et al.
Cycles in synchronous iterative voting: general robustness and examples in Approval Voting (2022) by Kloeckner
Decentralized Society: Finding Web3’s Soul (2022) by Weyl et al.
the top fifth [or top 10%, or whatever] of the proposals by the agreed upon scoring scheme [the scoring scheme is variable — i.e., maybe a proposal being ranked #2 gives 4 points and being ranked #5 only gives 1 point; maybe proposals ranked 1 to 3 are given 9, 4, and 1 points respectively, but those ranked 4 and 5 are given 0 points] are kept and voted on in the following round; the bottom 80% [or 90%, or 99%] are discarded along with those proposals made by accounts that did not participate in ranking proposals
the remaining proposals are ranked again by each of the 10,000,000 participants in a similar manner as before, with the bottom 80% [or 90%, or 99%] being disqualified; this repeats in succession until we are ultimately left with the desired number of “winners” that have been shown to be the most popular submissions among protocol participants
assuming 10,000,000 participants and proposals and the protocol keeping top 10% after each ranking/voting round, it takes the protocol 7 rounds to [though each “round” may involve that each participant ranks multiple sets of N proposals] whittle 10,000,000 proposals to 1 top proposal [10,000,000 x (0.1)^7 = 1]
assuming 8,000,000,000 participants and proposals and the protocol keeping the top 1% after ranking/voting round, it takes the protocol only 4 rounds to narrow 8,000,000,000 proposals into 80 proposals [8bn * (0.01)^4 = 80]; the last 80 proposals could be ranked by each participant in a final round, in which the top proposal [or top 5 proposals, or top 10, whatever] are kept as “winners”
the top fifth [or top 10%, or whatever] of the proposals by the agreed upon scoring scheme [the scoring scheme is variable — i.e., maybe a proposal being ranked #2 gives 4 points and being ranked #5 only gives 1 point; maybe proposals ranked 1 to 3 are given 9, 4, and 1 points respectively, but those ranked 4 and 5 are given 0 points] are kept and voted on in the following round; the bottom 80% [or 90%, or 99%] are discarded along with those proposals made by accounts that did not participate in ranking proposals
the remaining proposals are ranked again by each of the 10,000,000 participants in a similar manner as before, with the bottom 80% [or 90%, or 99%] being disqualified; this repeats in succession until we are ultimately left with the desired number of “winners” that have been shown to be the most popular submissions among protocol participants
assuming 10,000,000 participants and proposals and the protocol keeping top 10% after each ranking/voting round, it takes the protocol 7 rounds to [though each “round” may involve that each participant ranks multiple sets of N proposals] whittle 10,000,000 proposals to 1 top proposal [10,000,000 x (0.1)^7 = 1]
assuming 8,000,000,000 participants and proposals and the protocol keeping the top 1% after ranking/voting round, it takes the protocol only 4 rounds to narrow 8,000,000,000 proposals into 80 proposals [8bn * (0.01)^4 = 80]; the last 80 proposals could be ranked by each participant in a final round, in which the top proposal [or top 5 proposals, or top 10, whatever] are kept as “winners”
No activity yet