Attention Streams of BrightID
Public.icon
https://gyazo.com/393053ec322ccbd4c0c52956a024b841
Attention Streams
For an introduction to Attention Streams, please refer to this article.
Overview
Voting systems with economic effects are difficult because money can enter governance in undesirable ways. One actor can perpetuate rules that are disproportionately advantageous to themselves, but it is difficult to unite against them. When the proportional effect of voting on individuals is small, individuals may not be motivated to vote. Oligarchy and plutocracy are associated with the dysfunctional government types that arise from such systems, and Attention Streams aim to expose and defeat them.
It is difficult to get people to participate in governance if the rewards obtained by the proposal go to someone else. It is easier to just ride along with someone who is paying attention to that problem. In the economic realm, people are more likely to try to make changes if there is a possibility of making money.
The purpose of Attention Streams is to use the familiar strategy of making money in the economic realm, that is, finding popular ideas quickly and making them work in the governance realm. When people discover that they can make money in this way, their participation will lead to an increase in funding for public goods and better rules for themselves in economic systems. The right to vote in Attention Streams arises in a form that allows early discoverers to fix less capital than latecomers. Latecomers can receive rewards based on the attention paid to early discoverers.
Attention streams can be used for discovering and funding public goods, allocating rights to shared resources, drawing attention to proposals, and promoting idea generation. Attention streams can be used to determine rules for governments, DAOs, markets, and other attention streams. Let's look at an example.
Mechanism
Contributors help determine outcomes by acquiring shares of options that may turn into rewards, and continue to invest as long as that signal is needed. Funds to realize options are collected from contributors who support those options. Contributors who support popular options can pay rewards early.
Something like whether you can stake early on?tkgshn.icon*3
In contrast to traditional voting systems, where even impactful voting results have minimal value per vote, contributors can have a direct economic impact on their choices.
Comparison with existing technologies
Refer to the mechanism.
Attention streams are similar to registries (e.g. idea markets), prediction markets, futures markets, futarchy, and token-curated registries, each with their own bonding curve for each entry. Attention streams can be used to set rules for such systems. See examples.
Attention streams are similar to conviction voting in that users can allocate voting power among multiple options and voting is automatically strengthened over time. The difference is that funds can be obtained directly from voting actions, in addition to external sources, or instead of them, and that contributors can bet their own funds behind options. (This is consistent with public choice theory.)
Attention streams make early signaling more likely by allowing positions to grow automatically and then acquire a share of subsequent contributions. Because positions grow automatically, less capital is needed to maintain signals.
The Attention Stream sharing system makes it easy to apply external rewards to contributors in an organized manner. Late contributors to the option can consider rewarding early contributors, and external funding sources can fulfill this role. The Attention Stream can measure and rank support in a quadratic fashion. The economic difference between this and the bonding curve is that time is divided into cycles, and all contributors in the same cycle have the same opportunity, preventing early runs. The available tokens (not necessarily available for acquisition) increase over time and are reset at the end, making it highly incentivized to maintain a position. Because the benefits obtained in tokens are fixed, there is no reduction in returns for contributors even if others quit, and potential returns may increase instead. Long-term holders may also receive better returns than the bonding curve. In Simon de la Rouviere's concept of the curation market, voting tokens are bound to each curator's price curve. Curator reputation can be measured by the percentage of tokens bound to them. In contrast, in the Attention Stream, contributors can advertise their past performance and call on others to vote for them. Each Attention Stream arena can define a reputation source. The amount of contribution may be restricted by the reputation of the contributor. For more information on how Attention Stream posters build their reputation, please see here. Design for Attention Stream: Designers create arenas (rule boundaries) in a collection of topics (things to be governed or decided on) and create choices that are potential outcomes in the form of competing rules or decisions within the topic. Other designers can create new topics within the arena. Posters can add options to topics. Topics and choices must comply with the rules set by the designer, or they will face disputes. Use case examples provide specific examples of arenas, topics, and choices.
Options can earn tokens for contributors, or shares for representing initiatives. Arenas and topics can also have their own funds to reward designers and other initiatives. External funds can be applied to topic funds.
The number of tokens that can be applied to an option is limited by the token supply and sometimes by the reputation of the contributor.
Arenas, topics, and options cannot be taken down unless there is a conflict.
Arena Design
Designers determine the following for a new arena:
- Rules regarding the acceptable range of topics and options
- Ban on duplicate topics
- Any rule-breaking topics subject to dispute resolution framework
- Conflict-resolution framework for rule-breaking topics and options
- Tokens to be used
- Minimum contribution requirements (to protect against spam)
- Reputation system (if any)
- Correspondence between reputation and number of tokens contributed
- Funds
- Arena funds
- Location or address of the arena
- Fees (to protect against spam)
- Percentage of each contribution's share taken as fees for arena funds
- Fee for creating a topic
- Fee for creating an option
- Are topic funds permitted?
- Can they include external funds?
- Are option funds permitted?
- A tag describing the arena (useful for searching past actions of other posters by type of arena; see reputation)
Topic Design
Designers determine the following for new topics within an arena:
- Rules regarding the acceptable range of options and how they will be executed
- Ban on duplicate options
- Dispute resolution process for option-breaking rules
- How frequently shares are distributed (in other words, how long a cycle is) (e.g. every 24 hours)
- How many shares each position will receive per cycle (by percentage of tokens contained in the position)
- Are options executable?
- For how long a period of time must an option be at the top among competing options in order to be executed?
Funds
The percentage of contribution points paid to past contributors of a particular option. (Poster fee)
Topic fund (optional if allowed in the arena)
Location of the topic fund.
Percentage of each contribution taken as a fee for the topic fund.
Can the topic fund be regularly transferred to the option that has been consistently above the support threshold?
What is the relative support threshold?
How often is the topic fund transferred to the option that has been above the threshold for the longest period of time? This is called the funding period.
What percentage of the topic fund is transferred at each funding period?
Is choice funding allowed (if allowed in the arena)?
Tags that describe the topic. (These are useful for searching for past actions of other posters depending on the type of topic. See reputation)
Design the option
Description of the option
Compliance with arena rules and topic.
Must be clear enough to be enforceable as a result.
Indicates which other options it is competing with.
Failure to do the above will result in a dispute.
Funding for the option (if allowed in the topic and arena)
Location of the choice fund
Percentage of each donation taken as a fee for the choice fund.
Is there a target amount for the choice fund?
Once reached, the choice will no longer accept donations.
Limitations on fees
Arena fees must be less than 100%
Topic fee + contributor fee + arena fee must be less than 100%.
Choice fee + topic fee + post fee + arena fee must be less than 100%.
Mechanism
Contributors hold one or more positions (combinations of shares and tokens) in one or more options. Positions include tokens contributed and tokens earned from subsequent contributors. Contributors earn a share of the position each cycle. Withdrawing tokens from a position results in a loss of share.
Acquiring shares
Time plays a major role in Attention Streams. The cost to maintain support for options decreases (cost to add support increases) as positions earn a share continuously. Therefore, contributors will seek new opportunities rather than stacking what is already found. Topics may naturally disappear when the ranking of options is fixed.
In the example above, 10% of contribution goes to existing contributors of the option and shares occur at a rate of 100% per token per position cycle. (Token for position is equal to the difference between tokens contributed, fees paid, and fees earned.) These two parameters are set by the topic designer.
Contributor fees
Posters can earn tokens as fees from subsequent posters for the option. The contributor's fee is distributed proportionally based on the number of shares held by each contributor at the time of the new post. Tokens earned as a fee are automatically added to the contributor's position.
Pricing and measurement support
To compare options, a cost-per-percentage-point (CPP) is used to obtain the share of shares of options for the next cycle. For example, to obtain 1% point of the entire option of an option in the next cycle, it costs 1 token if the token is "big" and the option currently has 100 shares. (For simplicity, the number of shares added in the next cycle, including those from new donations, are not included in this calculation.)
This is displayed in the tokens that have been adopted for the topic.
If a topic is set up as a competition, CPP can be used to measure and rank support for options.
Relative support rates using funding from options of different scales
If options can set their own funding goals, it is effective to compare the relative support for the size of the funding goal. Relative support rates can be expressed as ratios (percentage).
Relative support = CPP / 1% of the selected funding target.
Secondary support
The mechanism of stocks and fees is suitable for measuring the support rate quadratically. In this case, support for an option is measured as the sum of the square roots of the share of each unique contributor. This inherits the sensitivity of quadratic voting to Sybil attacks and collusion, where attackers purchase or rent unique accounts to create the illusion of broad support. BrightID can be used to counter Sybil attacks, but MACI is currently impossible to counter collusion. In this framework, options must be known in advance, and the flow of attention is designed to discover options rather than determine known options.
Position cancellation
The number of tokens in a position is the number of contributed tokens minus the paid and earned fees. Tokens in a position are withdrawn by burning stocks. Tokens are withdrawn in proportion to the number of stocks burned. For example, if a contributor has 100 shares and 10 tokens in a position, burning 10 shares will withdraw 1 token. Tokens and stocks cannot be separated except by burning, but positions can be sold.
Position sale
The selected stocks come with tokens. The only way to separate tokens is to burn stocks. Stocks can be transferred (e.g. sold). In this case, a proportional number of tokens are transferred together from the position. For example, if a poster holds 100 shares and 10 tokens in a position, and transfers 10 shares to another person, that person also transfers 1 token. The 10 shares and 1 token are still attached to each other. Positions transferred in this way can be split between multiple wallets or recombined in a single wallet.
Cycle
The time is divided into cycles. The number of new shares generated in the next cycle is determined by the number of tokens in the position in the previous cycle. When tokens are added, fees to existing contributors are paid immediately, but only the share balance from the previous cycle is used. This prevents frontrunning.
Example
Let's see an example of how shares reward contributors.
In this example, 20% of the contribution is paid to contributors (proportional to the shares they hold), and shares are generated at 100% per token per cycle. Contributors A, B, and C are adding or withdrawing positions for the same option.
Cycle 1
Contributor A purchases 10 token worth of shares.
Contributor B purchases 10 token worth of shares.
End of Cycle 1
Shares
Tokens
Contributor A
0
10
Contributor B
0
10
CCP (Cost per share): 0.0
Cycle 2
Contributor A acquires 10 shares by having 10 tokens in his position.
Contributor B acquires 10 shares by having 10 tokens in his position.
Start of Cycle 2
Shares
Tokens
Contributor A
10
10
Contributor B
10
10
Contributor C purchases 20 token worth of shares.
(4 tokens go to the poster20%. 2 each go to contributor A and B) End of Cycle 2
Shares
Tokens
Contributor A
10
12
Contributor B
10
12
Contributor C
0
16
CCP (Cost per share): 0.2
Cycle 3
Contributors A, B, and C acquire shares according to the number of tokens they hold.
Start of Cycle 3
Shares
Tokens
Contributor A
22
12
Contributor B
22
12
Contributor C
16
16
Poster A adds another 10 tokens (2 tokens go to the poster 20%)(0.73 goes to poster A, 0.73 goes to poster B, 0.53 goes to poster C) End of Cycle 3
Shares
Tokens
Net Tokens
Contributor A
34
20.73
+0.73
Contributor B
34
12.73
+2.73
Contributor C
32
16.53
-3.47
CCP (Cost per share): 0.96
Cycle 4
Contributors A, B, and C acquire shares according to the number of tokens they hold.
Start of Cycle 4
Shares
Tokens
Contributor A
54.73
20.73
Contributor B
46.73
12.73
Contributor C
48.53
16.53
Contributor B burns 23.36 shares and withdraws 6.36 tokens (50% of the position).
Cycle 4 ends
Shares
Tokens
Net Tokens
Contributor A
54.73
20.73
+0.73
Contributor B
23.36
6.36
+2.73
Poster C
48.53
16.53
-3.47
CPP (Cost Per Share): 1.27x
Disputes
Any option that violates the rules of a topic or arena can be disputed at any time. If a dispute is successful, the option will be removed. Options have two sides: "dispute" and "don't dispute." Existing posters are automatically added to the "don't dispute" side. Posters have one cycle to post on either the "dispute" or "don't dispute" side, switch sides, or remain neutral. If there are no contributions left on the "dispute" side at the end of a cycle, the dispute is canceled with no effect. If there are no contributions left on the "don't dispute" side, the option is removed.
Disputes that include a topic (in place of an option) are handled similarly, with each side starting with zero contributors.
Staking
After one cycle for contributors to switch sides, all tokens on the losing side are considered "staked" on the outcome of the dispute resolution. An equal number of tokens from the winning side are also "staked." Any excess tokens are returned in reverse order of addition.
The option (or topic) then escalates to a more restricted economy (such as Celeste) for resolution.
All shares on the losing side are burned. The winner of the dispute receives the staked tokens of the losing side, divided proportionally among the staked tokens of each winning contributor (not share).
If the "dispute" side wins, the option (or topic) is removed.
Reputation
Contributors have wallets for providing and receiving funds. By reusing wallets, past actions can be linked. Wallet owners can incentivize others to add to their position by displaying past actions and rewards obtained from them.
All past actions and their economic outcomes are generally publicly visible from the wallet profile. If the user does not want to link past actions, they can use multiple wallets. On the other hand, users can also maintain the same wallet for a long period of time to gain reputation. Users are free to use their wallet to prove their identity outside the Attention Stream platform.
Users can optionally display social proof or highlight past actions on the profile page. Additionally, by browsing other users' past actions, users can narrow down and evaluate positions based on attributes such as earnings, topics, and arena types.
Restrictions on contribution by reputation
Arena designers can assign a reputation scheme and set limits on the amount of contributions that increase as reputation grows.
Use cases
Attention Stream can be used to discover and fund public goods, allocate rights to shared resources, get attention for proposals, and promote idea generation. Attention Stream can be used to determine the rules of government, DAOs, markets, and other attention streams.
Subjective ranking (ideas, art)
Case study SongADAO
Background and objectives
SongADAO supports Jonathan Mann's "Song-a-Day" activity. He plans to create 5,000 Song-a-Days by September 9, 2022.
SongADAO aims to increase listener engagement and provide a search tool for song registration. To achieve these goals, SongADAO is conducting a pilot attention stream to create subjective rankings of songs in various categories through crowdsourcing.
Token
The token used is the SONG token, which is a divisible token also used in the Song-a-Day NFTx pool. SONG represents ownership of songs in the pool, including fractional ownership, and can be obtained by purchasing on decentralized exchanges or by adding songs to the pool.
Fees (Arena, Topic)
The majority of the value of Attention Stream comes from listeners ranking songs, so fees for previous contributors should be high. Additionally, arenas will incur fees to go to SongADAO for artist support, pilot research and development, spam prevention, and more. Topics can also take a small fee to reward innovativeness of the category creator (see "Song Categories").
Song Categories
The ranking of songs is generally measured by support rate. However, there is not just one ranking for a song. Instead, the community can create new categories (=topics) for song rankings. For example, "most happy," "most sad," "most funny," "most folk," "most electronic," and so on can become completely different rankings. The category itself can also be ranked by the total support rate of all the options within it (and thus the potential accuracy).
Engagement
New categories have endless possibilities and can attract listener interest as a result. In order to perform well, the poster will need to listen to and carefully consider many songs. Because new categories are always being added, rankings are always fresh and latecomers may also get the chance to win.
Futarchy
The "metric" used in Vitalik Buterin's example of futarchy in his paper is purely monetary, but there is a clear way to measure it: GDP.
As suggested by Bo Waggoner, DAOs may want to measure success in ways other than the price of their own tokens.
The following is a sketch of how Attention Stream may be used to determine the metric and market (potential behavior) of futarchy:
In arena design, the metric to be used will become one topic. The description of the options must describe sufficiently how the metric will be measured so the oracle can follow it. A large portion (say 20%) of each contribution can go to previous contributors as a reward for innovation.
It is difficult to determine the appropriate weight for each metric. If a single snapshot of support for the choice of metric options is used, an attacker can show a significant contribution to the snapshot's date. Ideally, metric weights would be determined slowly and one at a time, giving contributors enough time to respond to attempts by wealthy contributors to control the outcome. One idea is to establish three (for example) metric frames with weights of 55%, 30%, and 15% each. The first frame is given to the first option with the highest support rate, which is maintained over a pre-set period (e.g. one week). The option is then removed (although it can theoretically gain multiple frames, as long as the support rate is reset and re-added), and the second and third frames are competed for in the same way.
The second topic is that futarchy can be used to select actions that futterkey considers. From the options with support above the threshold defined in the topic, a binary market for "yes" or "no" actions for the future image will emerge.
Funding for charities and public goods
Each topic represents a category of charity and can be treated as a single topic for all charities. Options raise funds from contributors who normally work in their arenas. Each option represents a charity. 10% (e.g.) of the donation is used to fund the option (i.e. charity). 10% (e.g.) is paid to past contributors.
Memes for charity
There is one topic that represents charity. Each option becomes a meme and competes with other memes to raise funds for charity. 20% (e.g.) of the donation is donated to the topic fund (e.g. charity). 10% (e.g.) is sent to past contributors.
Knowing how funds are used (e.g. for DAOs)
DAOs can secure an idea pool (topic fund) for discovering new proposals.
Here is an example workflow.
1. Use the attention stream to identify the proposal the community wants to see.
2. Proposals (choices) that win will be given funding-- the amount of which will be determined by Attention Streams.
3. We will recruit a team to work on the proposal that is selected.
4. The final winner will be chosen through a 2nd selection process by Attention Streams.
5. If the winning team (proposal) needs more funds than what they received from Attention Streams, they can apply for the remaining funds from the DAO.
By following this flow, even large proposals can be discovered and partially funded using a relatively small pool of funds set aside for idea creation through Attention Streams- for example, 10% the size of the Community Pool. The main Community Pool can supplement the remaining funds for partially-funded proposals.
Proposal selection
Refer to the External Funding Sources section for information on topic funding. Topics may fall under the category of public goods or proposals that the DAO has an interest in funding. Each month (for example), the topic fund may be supplemented with external funds (for example, from the Community Pool).
A contribution of 10% (for example) can be allocated to previous contributors. 10% (for example) can be allocated to selected proposals. 10% (for example) can be allocated to the topic fund for future winning proposals.
Regularly (for example, weekly), a percentage of the topic fund (for example, 50%) will be transferred to the choice (fund) that has exceeded the relative support threshold for the longest time. It is important to note that winning choices will only receive a portion of the topic fund (for example, 50%), and non-winning choices that have exceeded the relative support threshold will have an advantageous start during the next funding period even if no additional external funds are added. If none of the choices reach the relative support threshold, the topic fund will be carried over to the next transfer period.
If a funding option reaches its target amount, it will be removed from the topic. The added topic funding will be directed to the next options that meets the above condition. (This mechanism allows for funding multiple proposals of different levels simultaneously). In the next funding period, if more funds are needed, (contributions, shares and support will be reset to zero) the proposal can be re-added, or an application can be made to the DAO. This prevents valuable proposals from being blocked by low-value proposals for long periods of time.
Team Selection
A topic will be set for each winning proposal. Each team can apply as a selection. 10% (for example) of the post will be paid to past contributors. 10% (for example) will be topic funding and will be transferred to the winning team.
Completing Fundraising Using the DAO
The winning team must create a proposal (for example, a confident vote proposal) and apply for the necessary funds gathered in the flow of selecting the proposal and the team's choice in the spotlight of the proposal.
Selecting Rental Property Terms
Cf. Land Return.
Each plot of land (real estate is a public good) is a topic. The options within the topic represent parts of the lease agreement such as, the length of the period, the name of the lessee, etc. As some of the options are inevitably contradictory, this must be stated in the option description (otherwise it will result in controversy).
Arena Fund's fees are high to prevent spam proposals (for example: 20%). Arena Fund represents the community's collection of property usage fees, which can be used as a sociable dividend or to buy more properties (using another arena tuning to make the decision on how to use the expenditure).
Representing potential tenants as a set of opposing choices allows them to bid against each other, and community members can also weigh in economically. Those who have previously rented will have accumulated a share of those choices and may have an advantage over new choices. Shares added each cycle should be low enough to encourage healthy competition.
\n
5% (example) of contribution goes to previous contributors.
\n
In addition to setting contract terms, choices can also provide funding to improvement projects.
\n
In topic rules, a specified period in which the new choice must remain at the top must be set before making changes to the contract. For example, in the case of housing, a choice may be required to maintain its top position for six months to change part of the contract (including tenants).
\n
Topic rules also apply to contracts and cannot be overwritten by choices. This could include mineral rights or preservation of resources on the land. By carefully constructing topic rules, choices that could damage or steal public goods can become a focal point of conflict.
\n
Common topic rules when renting a property in one arena can be determined in a different (meta) arena.
\n
Discovery and funding of music and art.
\n
Each topic defines an aesthetic category (in its rules). 10% (example) of the post goes to past contributors. 10% of the post goes to the selection fund for artist payments. 1% (example) goes to the topic fund for designer payments.
\n
Economic considerations.
\n
Contributions against public welfare.
\n
There is a risk of contributors being deceived or bribed into submitting (or not opposing/dissenting) choices that represent false public goods.
\n
Attackers may also openly create harmful choices, threaten to make large donations to those choices in the future, and tempt others to donate ahead of them to earn fees.
Conflicts can be useful when a choice violates the rules. Choices with more funds or support behind them will naturally receive more scrutiny. A choice that conflicts with existing choices must either be explicitly stated in the description or cause controversy. This will automatically draw attention to good choices that counter harmful ones (or harmful choices that counter good ones). These new choices can be created to oppose the harmful ones and vice versa.
Latecomer\n\sMalicious\n\sShowing confidence in making a major future contribution can be seen as an attack, but it can also give opponents time to react and contribute to competing options or voice objections to them (see Donating to the Public Interest). If a tycoon tries to gain support by indicating future donations, early false donors can accept the tycoon's donations and withdraw their support before the choice is established. For more information on these settings, see Designing Topics.
Compassionate\n\sLatecomers who are compassionate can choose to apply matching funds to many choices in a topic. They can examine the choices and decide which ones to match and at what amount (optionally using 2nd matching). Attackers who expect matching funds may create false public goods or try to deceive the public with a civil account. Defenses against this are mentioned in secondary support.
Funding Topic with External Funds\n\sIn some cases, external funding sources added to a topic's fund may boost choices with additional funds. In this case, choices that exceed the relative support threshold the longest can receive additional funds from the topic fund, up to the amount requested by the choice.
In topics, clear rules are needed regarding which options are allowed to prevent harmful choices from being stolen from the topic fund. Since only one option can receive extra funds at a time, special attention will be paid to the promising options in advance. See the example of "Discovering the use of funds" for more information.
Attack methods include creating and tempting others to contribute to false public goods. Contributions to public goods are being discussed. Tempting others to contribute to the harmful choice or preventing opposition to it by fooling matching funds or other donors is possible if the attacker uses a Sybil account or bribes other donors. Defense measures against this are discussed in the Secondary Support section.
Paying someone to create a new option is useless. The attacker can create that option and get the same reward.
Paying someone to remove an option is not possible according to Attention Stream Design unless it is not controversial.
An attacker can do this themselves and does not need to pay anyone to support an option. The exception is to create a misconception of support when support is calculated particularly quadratically.
Posting a bribe to remove support for a competing option requires the poster to add support with another account after receiving the bribe.
If the attacker qualifies to receive a bribe (for example, by making the total support under the threshold at which a competing option is established), those who maintain support receive increasingly large contributor rewards from those who remove support and add their support with another account.
At some point in the future, indicate your intention to put money into the choice. This is being discussed in the section on major late donors. Loot the topic fund, as explained in the section on external funding for the topic fund. Frequently Asked Questions: Can you give an example of why an ordinary donor might be more advantageous than a wealthy one? For example, consider choosing an option with a total fee of 20%, with a poster fee of 10% and other fees of 10%, for 10,000 tokens as a wealthy (P) size bettor. If P puts 10,000 tokens into the selection and pays 2,000 tokens as a fee, to break even P needs an additional 20,000 tokens contributed by follow-up donors. It is easy for a regular participant (N) who puts in 100 tokens to break even; N just needs to add 200 tokens. For more on the Attention Stream, see this introductory article. Acknowledgments: Thank you to Aaron Soskin, Daniel Ospina, Spencer Graham, Rome Viharo, et al. for their reviews and comments. Auryn Macmillan suggested that 1. The first donor does not pay extra to the Arena Fund as compensation for not paying the previous donor, and 2. Price the choices from the perspective of acquiring some percentage of their stock. Vitalik Buterin's paper critiques Nathan Schneider's paper on the limits of crypto-economics and helps identify situations where the Attention Stream is useful. Glen Weyl at Radical Xchange provided lots of inspiration. Jonathan Mann provided a use case for SongADAO. Thank you to everyone who commented on this document.