Recommended
Recommended
More Related Content
Similar to Better Information with Curation Markets
Similar to Better Information with Curation Markets (20)
Recently uploaded
Recently uploaded (20)
Better Information with Curation Markets
- 1. Better Information w/ Curation Markets Spencer Graham — @spengrah
- 2. Spencer Graham — @spengrah Finding signal in the noise is hard Finding signal in the noise = curation 1. Find the data 2. Organize it
- 3. Spencer Graham — @spengrah How we find data impacts how we can use it 1. How distributed is the data? 2. How related are… ○ the incentive to find the data, and ○ the structure of the data?
- 4. Spencer Graham — @spengrah For example... Data distribution Data structure(s) Incentive to contribute Match? Result Articles about everything distributed Narratives Network graph Reputation; “civic duty” ✘ Wikipedia needs to provide infrastructure and maintenance Central point of failure Social profiles distributed Arrays of attributes Network graph Connect with your friends; build your brand ✘ Facebook/Twitter/LinkedIn needs to provide infrastructure and see outsized profits and power Value of scarce resources distributed Prices Money ✓ Spontaneous curation!
- 5. Spencer Graham — @spengrah DistributedConcentrated Less related More related How distributed is the data? How related are the incentive to find the data and the desired data structure? Wikipedia Kaggle Prices Reddit upvoting Facebook Some examples today
- 6. Spencer Graham — @spengrah Find: already found Organize: centralized actor needed Some information lives here: ● Some objective information ● A single person’s subjective opinion ? Find: centralized actor needed Organize: centralized actor needed Majority of information lives here: ● Most objective information ● Most subjective information Spontaneous curation Permissionless contribution with a direct incentive to contribute data in desired structure No centralized actor needed Little information lives here: ● value of scarce resources DistributedConcentrated Less related More relatedHow related are the incentive to find the data and the desired data structure? How distributed is the data? Most data is locked up, often inside centralized organizations
- 7. Spencer Graham — @spengrah PricesNarratives Curvilinear relationships Databases - relational, dimensional, nosql Network graphs Trees Ranked lists (interval or ratio) Ranked lists (ordinal) Unranked lists (sets) In terms of data structures... Distributeddata Less related More relatedHow related are the incentive to find the data and the desired data structure?
- 8. Spencer Graham — @spengrah Spontaneous curation with tokens TCR image from Sebastian Gajek. Bonding curve image from Slava Balasanov. tokenize the data structure → token-curated registry tokenize the data itself → curation market Discrete set membership Continuous values
- 10. Spencer Graham — @spengrah What is a token-curated registry? A list of items... curated by a community... coordinated by a dedicated token. A discrete membership data structure... produced organically by a set of individuals... each maximizing their token holdings.
- 11. Spencer Graham — @spengrah Example: good bars in ChicagoHow do TCRs work? 100 bar tokens 100 bar tokens Bar owners apply() challenge() Bar critics/ enthusiasts Adapted from Sebastian Gajek. good bars in Chicago TCR bar status Sheffields ✓ Billy Sunday ✓ Mad River ✘ Rainbo Club 3 types of actors: ● Applicants ● Community Curators ● Consumers Bar goers vote() 10 bar tokens free to read
- 12. Spencer Graham — @spengrah TCRs work when... ● Registry objectives are clear ● Curators have an incentive to maintain list quality ● Token value is increasing ● Consumers value the information from the list ● Potential applicants value being part of the list ● There is increasing token demand from applicants to pay for application fees/staking ● Voting functions as expected
- 13. Spencer Graham — @spengrah TCRs may fail when... ● The maximum economic value of the registry is not high enough → application stagnation ● Curators don’t have enough bandwidth to sufficiently evaluate applications ● Token holders are passively speculating or freeriding ● There is no dedicated token (DIRT protocol is testing this hypothesis) ● Smart contract devs can’t iterate on the token mechanics ● They can’t stand up to adversarial attacks ○ 51% attacks ○ Collusion ○ Bribery, including the super-scary Dark DAO problem ○ Griefing
- 14. Spencer Graham — @spengrah TCR voting ● Risk of herding behavior if voting rewards are based solely on consensus ○ Possible misalignment with registry quality and token value ○ Also risks vote-then-exit behavior ○ Vary the reward and penalties by vote distribution? ● Objective vs. subjective information ○ Higher risk of herd voting for subjective topics? ○ Objective curation involves more work; is a higher incentive necessary? ○ If curators expect cheating to be rare, there’s a lower incentive from challenge rewards. Might need to introduce forced errors. ○ Break up information into smaller more easily/objectively evaluated bits (framework-based TCRs) ● What if voting behavior isn’t rational? (humans be humans) ● Are token reward incentives sufficient? Do we need a form of reputation or “knowledge”?
- 15. Spencer Graham — @spengrah TCR token engineering design space Application fee ● Magnitude ● Currency Voting incentives ● Rewards ● Slashing ● Forced-errors Voting mechanics ● Quorum requirement ● Listing default ● Commit-reveal (PLCR) ● Secret voting ● Delegation ● Knowledge requirement Block rewards / inflation Bootstrapping ● ICO ● Continuous token
- 16. Spencer Graham — @spengrah - Videos - Transparent ICO projects - Physicians - Quality “newsrooms” - Marketplace/community whitelist - Quality online publishers - Non-spam artworks Unordered Data structure: unranked list / set Registry design patterns Lots more
- 17. Spencer Graham — @spengrah Ordered Data structure: ordinal ranked list Graded or Staked Data structure: interval or ratio ranked list Registry design patterns From TCR Design Patterns by Matt Lockyer From Graded Token-Curated Decisions by Sebastian Gajek
- 18. Spencer Graham — @spengrah Combinatorial Data structure: unranked list; narrative? Layered Data structure: ordinal ranked list Registry design patterns From The Layered TCR by Trent McConaghy From TCR Design Patterns by Matt Lockyer
- 19. Spencer Graham — @spengrah Nested Data structure: network graphs; trees Registry design patterns From TCR Design Patterns by Matt Lockyer
- 20. Spencer Graham — @spengrah Prices Network graphs: nested TCRs Tree: nested TCRs Ranked lists (interval or ratio): staked or graded TCRs Ranked lists (ordinal): ordered TCRs, layered TCRs Unranked lists (sets): unordered TCRs Narratives: combinatorial TCRs? Narratives Curvilinear relationships Databases Distributeddata Less related More relatedHow related are the incentive to find the data and the desired data structure? When we tokenize the data structure (TCRs)
- 21. Curation Markets
- 22. Spencer Graham — @spengrah Continuous Token Model ● Stake a currency (e.g. ETH) to a contract to you new tokens proportional to your stake ● ETH is now “bonded” to the contract ● The contract mints itemTokens and transfers them to the staker Curved Bonding ● A market-maker contract ● Price of itemToken is a function of its supply ● Can have different curves for buy and sell Some prerequisites ETH You Item Contract itemTokens “The anti-ICO” Bonding curve image from Simon de la Rouviere
- 23. Spencer Graham — @spengrah How do curation markets work? 1. Stake currency (e.g. ETH) to an item 2. Receive itemTokens proportional to your stake, as defined by the bonding curve 3. As others buy and sell, the price in ETH of an itemToken increases and decreases along the bonding curve When items using the same curved bond function, the market value of those items can be compared. Sheffields 20 ETH Billy Sunday 25 ETH Mad River .05 ETH
- 24. Spencer Graham — @spengrah Well, maybe… ...though some disagree. But not if we tie itemTokens to measurable utility ● Governance over the item ● TCR participation (e.g. a decentrally-curated cocktail menu) ● Coupons for the good or service Wait, isn’t this just a Ponzi scheme?
- 25. Spencer Graham — @spengrah Curation markets may fail when... ● The token does not have utility outside of its meme value ○ Though maybe meme value is sufficient? ● The token mechanics and agent behaviors are not well understood ● When the smart contracts are not upgradable ○ Enable iterative learning and adjustments ○ But also increases the attack surface
- 26. Spencer Graham — @spengrah Curation market design patterns Standard Data structure: interval or ratio ranked list A platform for oracles ● Anybody can create an oracle using the protocol ● Users interested in querying an oracle must bond the Zap token to receive dot tokens ● Each dot token is redeemable for one query of that oracle A good explanation by newsbtc.com
- 27. Spencer Graham — @spengrah Curation market design patterns Nested Data structure: trees and network graphs Sheffields 20 ETH Billy Sunday 25 ETH Mad River .05 ETH Appletini 20 BST Negroni 300 BST Old Fashioned 100 BST
- 28. Spencer Graham — @spengrah Curation market design patterns Proofed Curation Market Data structure: interval or ratio ranked lists Goal is to tie together... 1. Predicted popularity: via a standard curation market 2. Proofed popularity: how often a service was used Stakers to a service receive block rewards when they successfully provide that service to others ● Dataset providers (e.g. Alice) stake Ocean tokens to their dataset , receiving “drop” tokens per a bonding curve ● If a user downloads the dataset from Alice, Alice receives block rewards in proportion to the number of drops she owns for that dataset ● Anybody else can also stake to that dataset and make it available for download Read more from Trent McConaghy A decentralized data exchange protocol
- 29. Spencer Graham — @spengrah Prices Network graphs: nested TCRs, nested curation markets Tree: nested TCRs, nested curation markets Ranked lists (interval or ratio): TCRs with item staking, curation markets, proofed curation markets Ranked lists (ordinal): ordered TCRs, layered TCRs Unranked lists (sets): unordered TCRs Narratives: combinatorial TCRs? Curvilinear relationships: nested TCRs, curation markets Narratives DatabasesDistributeddata Less related More relatedHow related are the incentive to find the data and the desired data structure? When we tokenize the data itself (curation market)
- 30. Spencer Graham — @spengrah Conclusion? ● By tokenizing data structures, we create a direct incentive to contribute data, leading to spontaneous curation ● That could wrest the gains from curation from centralized organizations who build huge moats around their data assets ● With TCRs and curation markets, we have tools necessary to start tokenizing many types of data structures ● BUT, there are a lot of questions about that need to be addressed
- 31. Curation Markets = Spontaneous Curation