Points of Leverage in Improving Social Media Integrity Outcomes

Note: Listed in descending order of both increased leverage/efficacy AND difficulty to influence

12. Constants, parameters, numbers

• VM weights
• Engagement models
• Guardrails + regression budgets

11. The sizes of buffers and other stabilizing stocks, relative to their flows

• Human Capital Resources for building enforcements, labeling, moderating, fact checking etc etc
• Scale and scope of content needing to be evaluated

10. The structure of material stocks and flows

• Number of engineers + XFN resources working on Integrity
• Number of labelers/operations staff
• Survey pool constraints

9. The lengths of delays, relative to the rate of system change

• Improve long term experimentation tools/culture
• Improve survey based methodologies and feedback delays
• Model refresh cycles too slow to adapt to constantly changing cultural contexts/threats etc
• Delays in time needed to approve and rollout integrity changes vs the same for relevance

8. The strength of negative feedback loops, relative to the impacts they are trying to correct against

• Demotions are heavily capped, while boosts are uncapped/exponential
• Reduce effects or virality
• Number of engineers + XFN resources working on Relevance compared to integrity

6. The structure of information flows (who does and does not have access to information).

• Cooperative integrity efforts and data sharing
• Transparency with researchers/external stakeholders
• Rate of information sharing on platform - Repeat offender enforcement. and re-share friction
• Organizational structure of teams working on integrity problems (and relevance)

5. The rules of the system (such as incentives, punishments, constraints)

• Taking ownership of the VM
• PSC cycles and individual incentives (e.g. working on lower risk projects because of legitimate concerns about being blocked by policy, not working on longer term vision projects because of short term incentives)
• Content demotion legitimacy guidelines
• Policy influence on decision making
• Country carve outs for experimentation on Civic

4. The power to add, change, evolve, or self-organize system structure. Taking ownership of the VM (and objective function)

• Ability for integrity product to make more single threaded decisions vs massive cross-FB + senior leadership buy in

3. The goals of the system.

• Internal organizational goals/objectives.
• Objective function of the VM (which is influenced/determined by internal org goals)
• Ability to measure "good" for integrity/sensitive content vs only reducing "bad"
• Ability to measure "good" for user value beyond log behavioral engagement methods
• Feed optimizes for ecosystem aggregate values vs user level values

2. The mindset or paradigm out of which the system - its goals, structure, rules, delays, parameters - arises.

• FB business model
• FB culture
• Inherent challenges in balancing free speech vs enforcing on integrity, and meeting diverse international cultural norms
• Fiduciary duty

1. The power to transcend paradigms.

• Governance
• Regulation
• Employee Bargaining/Organization