Source reconstruction · Proposed
Evolving Prosocial AI
From Paul Atkins and David Sloan Wilson’s Evolving Prosocial AI · reconstructed by AI
Question 4 of 6 · The change
What will we change?
The change, and what it should lead to.
- Statements
- 25
- Connections
- 20
- Open questions
- 1
How to read it. Read from the bottom up: from each change to the effects it should have.
ChangePredicted effectDesired effectPossible harmObservationandJoint conditionsIs meant to preventFrom another part
Part 1 of 4
Make prosocial AI pay
Design principles, procurement, measurement and visible energy costs could turn prosocial AI from a handicap into an advantage. The authors' own caution about imposing principles is kept as a risk.
Draws on Paul Atkins and David Sloan Wilson (document).
Outline
- Predicted effectProsocial AI development gains an advantage rather than facing a competitive handicap.From Paul Atkins and David Sloan Wilson
- Possible harmApplying evolutionary, Western-derived principles overrides local wisdom, echoing how evolutionary ideas were once used to justify colonial harm.From Paul Atkins and David Sloan Wilson
- ObservationOstrom's research found farmer-governed irrigation systems in Nepal outperformed agency-managed ones; the authors cite it as cooperation made rewarding by institutions rather than virtue.From Paul Atkins and David Sloan Wilson
Beyond this part · 10 links
- Prosocial AI gains an advantage, not a handicap. contributes to Success at each scale reinforces the others. in “How prosocial worldviews spread”
- Apply the core design principles to AI. is meant to address Institutions reward engagement and short-term growth. In the way
- Apply the core design principles to AI. is meant to address Optimise for engagement, as competitors do. Pulling apart
- Apply the core design principles to AI. is meant to address Design for capability, even at a cost to engagement. Pulling apart
- Make procurement require capability audits. is meant to address Institutions reward engagement and short-term growth. In the way
- Make procurement require capability audits. put into practice by Procurement, funders and networks back prosocial AI. First steps
- Measure capability impacts across scales. is meant to address Immediate rewards outweigh delayed harms. In the way
- Measure capability impacts across scales. put into practice by Agreed capability measures. First steps
- Report energy costs and favour efficient models. is meant to address AI's ecological costs are externalised and ignored. In the way
- Learn from communities that already govern commons. put into practice by Tests in culturally different settings. First steps
Part 2 of 4
Build capability, not dependence
AI that scaffolds thinking and honours stated preferences could leave people more capable and more demanding of good tools. Mixed evidence is kept as a risk.
Draws on Paul Atkins and David Sloan Wilson (document).
Outline
- Predicted effectPeople build competence and autonomy and come to demand tools that keep building them.From Paul Atkins and David Sloan Wilson
- Possible harmAI that sharpens individual thinking disrupts the shared mental models a team relies on.From Paul Atkins and David Sloan Wilson
- ObservationThe authors report that handing writing to AI left them remembering and owning their arguments less, and that a slower, critical practice of drafting with Claude kept their capability.From Paul Atkins and David Sloan Wilson
Beyond this part · 5 links
- People grow capability and demand better tools. contributes to Success at each scale reinforces the others. in “How prosocial worldviews spread”
- Design AI to scaffold thinking, not replace it. is meant to address AI is built to capture attention and do tasks for people. In the way
- Design AI to scaffold thinking, not replace it. is meant to address Let AI complete the task. Pulling apart
- Design AI to scaffold thinking, not replace it. is meant to address Do the thinking themselves. Pulling apart
- Honour users' stated preferences over engagement. is meant to address AI is built to capture attention and do tasks for people. In the way
Part 4 of 4
How prosocial worldviews spread
Demonstrations, networks and stories spread prosocial worldviews through ordinary social learning. With the other changes, success at each scale could reinforce the rest.
Draws on Paul Atkins and David Sloan Wilson (document).
Outline
- Predicted effectProsocial worldviews spread through ordinary social learning as respected peers adopt them and report better outcomes.From Paul Atkins and David Sloan Wilson
- Desired effectSuccess at each scale reinforces success at the others: the life pattern rather than the cancer pattern.From Paul Atkins and David Sloan Wilson
- ↑ contributes to all of these together
- ↑ contributes to
- ↑ contributes to
- Possible harmMovements have to use engagement-maximising platforms to advocate for alternatives to them.From Paul Atkins and David Sloan Wilson
Beyond this part · 6 links
- Prosocial worldviews spread through ordinary social learning. contributes to Prosocial AI gains an advantage, not a handicap. in “Make prosocial AI pay”
- Build and publicise demonstrations of multilevel benefit. is meant to address Prosocial alternatives start at a disadvantage. In the way
- Build and publicise demonstrations of multilevel benefit. put into practice by Rigorous demonstrations at scale. First steps
- Grow networks where prosocial practice is experienced. is meant to address Training teaches extraction as common sense. In the way
- Grow networks where prosocial practice is experienced. put into practice by Procurement, funders and networks back prosocial AI. First steps
- Tell stories that make cooperation feel natural. is meant to address The worldviews shield one another from doubt. In the way
Where this comes from
In the Logical Thinking Process this view is a future reality tree. How a commons works →