Manifold Grace

Related research and venture roadmap

Working roadmap · v0.1 · 15 November 2026

These candidate studies sit alongside Integrated Societal Intelligence. Each can be developed within its own discipline or programme as evidence and project capacity allow.

Integrated Societal Intelligence roadmap · Automated venture ideas

Candidate studies and pilots

Agent reliability, reasoning faithfulness and synthetic training

AI evaluation · Study design in development

Paper or output. Technical paper candidates Project. Run separate controlled agent, reasoning-intervention and recursive-training experiments. Evaluation. Score error recovery, answer changes under matched interventions and rare-case retention.

Read-only outline · Long-horizon agent reliability · Read-only outline · Faithfulness of generated reasoning · Read-only outline · Recursive synthetic training and loss of coverage

Scarcity framing and network spillovers

AI evaluation · Study design in development

Paper or output. Behavioural economics and econometrics paper candidates Project. Design the framing experiment and a separate simulated network-policy comparison. Evaluation. Check whether AI-inferred links alter effect estimates and whether task wording changes measured beliefs.

Read-only outline · Scarcity framing and shared resources · Read-only outline · Policy effects under network interference

Algorithmic pricing and endogenous measurement

AI evaluation · Study design in development

Paper or output. Industrial organisation and measurement paper candidates Project. Use simulated pricing panels and performance metrics with known generating mechanisms. Evaluation. Measure false positives under common shocks and whether a higher metric score masks worse outcomes.

Read-only outline · Identifying algorithmic pricing patterns · Read-only outline · Endogenous measurement and Goodhart effects

AI adoption and early-career skill formation

AI evaluation · Study design in development

Paper or output. Labour economics paper candidate Project. Map junior tasks to learning opportunities in one profession. Compare unaided and supervised AI-assisted exercises. Evaluation. Assess later unassisted performance. Treat historical adoption and long-run employment effects as separate evidence questions.

Read-only outline · AI adoption and early-career skill formation

Private compute infrastructure and cross-border contracting

AI evaluation · Study design in development

Paper or output. Global governance and commercial law paper candidates Project. Trace one public compliance rule through provider records and build fictional automated transactions with varying authorisations. Evaluation. Distinguish documented conduct from inferred motives and separate execution location from authorisation.

Read-only outline · Private compute infrastructure and implementation authority · Read-only outline · Attribution in automated cross-border contracting

Disruption attribution and system-wide shocks

AI evaluation · Study design in development

Paper or output. Governance and methods paper candidates Project. Compare public post-incident findings and retrospective counterfactual designs using published or fictional records. Evaluation. Check timelines, alternative explanations and unsupported attribution. Keep the work focused on evidence assessment after an incident.

Read-only outline · Attribution after cross-border disruption · Read-only outline · Counterfactual inference after a system-wide shock

Local heat adaptation and social value procurement

AI evaluation · Study design in development

Paper or output. Climate governance and procurement paper candidates Project. Trace one heat commitment per borough and digital skills pledges in comparable contracts. Evaluation. Check claims against records of commitment, delivery, reach and outcomes.

Read-only outline · Equity and local heat adaptation · Read-only outline · Verifying social value in digital skills procurement

Oxfordshire procurement working paper · Public procurement evidence gaps project

Managed retreat, compound climate risk and fiscal stress

AI evaluation · Study design in development

Paper or output. Climate finance and governance paper candidates Project. Model one jurisdiction and one route from physical loss to fiscal exposure. Compare one-way and coupled scenarios. Evaluation. Verify extracted assumptions, test sensitivities and distinguish simulated results from observed impacts.

Read-only outline · Managed retreat and municipal finance · Read-only outline · Compound climate risks and fiscal stress

Transboundary climate intervention governance

AI evaluation · Study design in development

Paper or output. Climate governance and legal theory paper candidate Project. Use published assessments to compare evidentiary thresholds for investigation, precaution and compensation. Evaluation. Score how accurately AI conveys uncertainty and separates modelled effects, event attribution and legal responsibility.

Read-only outline · Transboundary climate intervention governance

Audit costs and market entry

AI evaluation · Study design in development

Paper or output. Regulatory economics paper candidate Project. Compare fixed costs and detection coverage under uniform and risk-tiered requirements. Evaluation. Use matched fictional audit cases and score errors detected, reviewer time and estimated compliance costs.

Read-only outline · Audit costs and market entry

Environmental credit integrity

AI evaluation · Study design in development

Paper or output. Environmental measurement paper candidate Project. Study one credit type with differing baseline, measurement and durability assumptions. Evaluation. Check whether AI distinguishes observed physical change from evidence of additionality.

Read-only outline · Environmental credit integrity

Emergency authority and democratic review

AI evaluation · Study design in development

Paper or output. Political science and institutional design paper candidate Project. Compare fixed expiry, renewal and retrospective review in fictional emergency decisions. Evaluation. Test authority reconstruction, missing reasons and renewal flags.

Read-only outline · Emergency authority and democratic review

Automated monitoring offers and pilot

Venture pilot · Proposed

Paper or output. Venture development Project. Use the existing 50-idea Venture catalogue. Develop a four-week digital skills tender pilot from a retrospective sample; retain parliamentary and permit monitoring as alternatives. Evaluation. Check source coverage, matching and correction rates, owner review time and willingness to pay.

Public Gap Register · Read-only manuscript outlines