Social Systems
| Social Systems | |
|---|---|
| Description | How to design/foster social structures / incentives / programs for fostering synthesis-friendly work, such as writing living literature reviews, adding metadata to experiments, etc.? |
| Related Topics | Incentive Systems |
| Discord Channel | #social-systems |
| Facilitator | Sílvia Bessa |
| Members | Jay Patel, Matt Clancy, Angelina Lesnikova, Valerii Kremnev, Raphael Walker, Martin Karlsson, Nouran Soliman, Sílvia Bessa |
What
How to design/foster social structures / incentives / programs for fostering synthesis-friendly work, such as writing living literature reviews, adding metadata to experiments, etc.
Async: Potential Next Steps
On "organize thought and writing on a central theme/problem to facilitate future work" track, if there is interest, I would be keen to bring together those that would like to map out:
- The structural options for rewarding contribution.
- Learnings from past and ongoing attempts (what truly motivates sustainable participation).
- New funding sources that could sustain those reward mechanisms.
Then map to the enabling effects of the tooling and model initiatives occurring across the workshop. One goal would be to connect the thread from these new funding sources and methods to the attribution and reward mechanisms offered by discourse graph and synthesis tooling.
Day 1: Goal for the workshop
Expected outcome
Resources, such as a system map/synthesis of the problem space, synthesis/directory of tools, essential reading list, case study library, or shared synthesis benchmark dataset
- The best tools in the world mean nothing if no one adopts them. Let’s merge together small scale testing, with rigorous validation.
- Create a resource that the tool builders across the group can glance at to not lose sight of the critical point of adoption. With practical examples of how these mechanisms are designed, iterated upon, and how they can be tested and communicated to academia.
- Learnings from past and ongoing attempts (what truly motivates sustainable participation).
- "A tool builder should check their assumptions agains this checklist before ..."
- validate assumptions for the system that should work on a bigger scale / models to test the incentives
- Example:
- Reputation & interactions in online space: how can we build better tools to help users navigate these spaces without being afraid of their reputation
Identified Open Problems
- Acceptance and onboarding of scientists, even if we have a model that works in a small setting
- Value attribution:
- How do you distribute rewards?
- Opportunity side: new tools looking at ways to provide input to that distribution mechanism
- How do we connect the two sides?
- How do we test the behavior of a model as it scales?
- How do we predict the inventive structures or perverse behavior that will arise as it’s adopted for a large number of players?
- Incentive mechanisms for contributing and maintaining a collective knowledge graph
Day 2: Outcome - Patterns and ID cards for new incentives mechanism design and adoption
ToDo: add summary /context for converging on Patterns and Incentive Mechanism Cards
❓Question What are the patterns for designing a new incentive mechanism system, from inception to adoption by academia?
❓Question How do we know if a new incentive mechanism or tool is ready for academia?
❓Question What drives the acceptance and onboarding of scientists?
💡 Claim Patterns
Patterns come from our reflective observations about our hands-on interactions to solve problems in the world. One simple definition is that a pattern is a repeatably applicable solution to a problem that arises in a specific context. What’s the pattern form?
See the example of the Scrum Patterns book that used "A pattern language: Towns, buildings, construction." book as reference
- Common building blocks of mechanism design:
- Problem
- Frequent approaches or solution
- How to test / quantify?
Any people that designs new mechanisms
ToDo: summarise the idea better
💡 Claim ID Cards for new incentive mechanisms
Serve to disclose information about a new mechanism design. What would be the sections of a model card for a new incentive mechanism? Who proposed it, assumptions made, type of incentivized behavior, riks and perverse behavior, evaluation, theoretical support…
ToDo: copy the 1st draft of Mechanism Cards from: https://docs.google.com/document/d/1g1V4WKKmIaEajC9AW9fdbqOnTd7joaasc-AjFDYNC5o/edit#heading=h.pjvcgnypdp3a
💡 Claim Wiki of incentive mechanisms
- Create an “incentive mechanism card”, a structure on the wiki
- imagine we have a wiki of all incentives ever designed, what would be the “mandatory” sections?
- Convert one of the examples in the chat to that structure
Both Patterns and Incentive Mechanism Cards can be converted to wiki structures. These resources are complementary in the sense that Patterns can be small cards that we consult when designing a new mechanism, while ID cards would be unique identifiers of new mechanisms (they can
Future Directions:
- Create a Wiki Structure for Incentive Mechanism Card
- Create and fill wiki pages for some examples of Incentive Mechanisms
- Idea: invite designers to create Patterns for Mechanism Design
Examples of Incentive Mechanisms
ToDo: copy from messages on Discord
Bibliography & Resources
- SourceCred
- Golden - Reward function development
- HyperCerts
- Impact Evaluators
- An Engine of Improvement for the Social Processes of Science, by Nielsen and Qiu
Just on giant labs book: https://www.biorxiv.org/content/10.1101/2022.04.25.489391v1.full.pdf
Other examples of pattern libraries: https://decentpatterns.xyz/library/
rescognito.com for contributions
Something built on top of https://credit.niso.org/
https://anagora.org is the prototype :)