Changes

Paul Atkins and David Sloan Wilson’s Evolving Prosocial AI across all six views: 105 proposed statements and 88 proposed connections, each with a page reference. No endorsement, trial or outcome is recorded.

• Added
  AI strengthens cooperation and flourishing at every scale.People's competence, autonomy and relatedness grow.Those who build AI share its benefits and a say.Democracy, shared facts and institutions get stronger.AI works within planetary limits.Cooperation pays better than extraction.Those shaping AI hold prosocial worldviews.Affected communities agree what AI is for.Contributors share benefits; costs aren't pushed onto others.Those affected help make the rules, including not to build.AI is audited, and violations meet escalating consequences.People can challenge harmful deployments and get remedies.Stakeholder bodies have real power, not just a say.Nested governance coordinates across scales.Capability impacts can be measured.AI scaffolds thinking rather than doing the work.Users' stated preferences are respected and can override.AI supports human connection rather than replacing it.People have the conditions to turn access into capability.Information systems favour agreement over engagement.AI helps citizens deliberate rather than deciding for them.AI is designed to widen opportunity for the marginalised.AI's energy and material costs are visible and weighed.Communities with ecological knowledge keep authority.Worldviews spread and persist through what they reward.Human cooperation is a partial evolutionary transition.Machine behaviour carries and spreads worldviews.Four extractive worldviews dominate AI development.The worldviews shield one another from doubt.Immediate rewards outweigh delayed harms.Institutions reward engagement and short-term growth.Power concentrates; alternatives are bought or lobbied away.Training teaches extraction as common sense.AI is built to capture attention and do tasks for people.Competence, autonomy and relatedness erode.Eroded users depend more and demand less.Firms that externalise harm outgrow the rest.Workers, creators and the knowledge commons go unpaid.The epistemic commons breaks down.Automated public systems harm people unaccountably.AI amplifies existing inequality.AI's ecological costs are externalised and ignored.Weakened institutions can't govern what weakens them.A social trap locks extraction in.Prosocial alternatives start at a disadvantage.An AI organisation lasts and does good.It needs to keep up with competitors.It needs to avoid harming users and society.Optimise for engagement, as competitors do.Design for capability, even at a cost to engagement.A person gets their work done and flourishes.They need to get things done quickly.They need to keep and grow their own capabilities.Let AI complete the task.Do the thinking themselves.Apply the core design principles to AI.Design AI to scaffold thinking, not replace it.Honour users' stated preferences over engagement.Spread cooperative ownership, open models and benefit sharing.Use AI that surfaces agreement across differences.Measure capability impacts across scales.Make procurement require capability audits.Report energy costs and favour efficient models.Build and publicise demonstrations of multilevel benefit.Grow networks where prosocial practice is experienced.Tell stories that make cooperation feel natural.Learn from communities that already govern commons.Prosocial AI gains an advantage, not a handicap.People grow capability and demand better tools.Citizens share facts and deliberate across differences.Contributors share in AI's value.Prosocial worldviews spread through ordinary social learning.Success at each scale reinforces the others.Imposed principles override local wisdom.Helpful AI weakens a team's shared understanding.Bridging fact-checks arrive too late.Movements must use the platforms they oppose.Farmer-governed irrigation outperformed agency systems.vTaiwan informed over twenty national laws.The authors changed how they used AI after losing ownership.Prosocial patterns take hold before lock-in.No agreed measure of capability impact.Agreed capability measures.Evidence skews Western and democratic.Tests in culturally different settings.Research depends on industry funding.Independent research funding.Extraction commands far more resources.Procurement, funders and networks back prosocial AI.Training reproduces extractive assumptions.Education teaches multilevel and capability thinking.Many prosocial models are still aspirational or niche.Rigorous demonstrations at scale.Enhancement and extraction look alike.Researchers develop capability measures.Measures can tell enhancement from extraction.Most AI tools do tasks and chase engagement.Technologists build scaffolding designs.Capability-building tools exist to compare.Buyers reward capability and extraction alike.Buyers require capability audits in procurement.Prosocial AI wins public contracts.Prosocial successes are little known.Funders back and publicise demonstrations.Prosocial approaches gain credibility.and 88 connections