Our research programme

Agency, meaning and the conditions of continuation

How do capacities to act become possible, gain recognition and remain sustainable—and who can change the conditions on which they depend? This question connects EdgeLab’s work on education, generative AI, institutions, political economy and everyday life.

The programme develops autopoietic ecology through theoretical writing, collaborative studies and exploratory design. Its recent work connects three concerns: how an action becomes possible or authorised, what sustains the capacity to act again, and how an acknowledged difficulty can change a subsequent decision.

Programme updated 18 September 2026. Project descriptions distinguish published research, submissions, working manuscripts and proposed investigations.

From teacher learning to ecological questions

Steven Watson’s earlier research on mathematics teacher learning, self-efficacy and professional judgement examined how people develop the capacity to act within particular settings. Work on educational politics, public authority and social systems extended that inquiry to the categories through which actions become recognisable and legitimate. Meaning mediation and autopoietic ecology bring these questions together while keeping lived experience, institutional communication and technical operations distinct.

The programme’s current directions

1. Autopoietic ecology, rough closure and hyperknots

The revised Autopoietic Ecology book with Erik Brezovec distinguishes the conditions that make an operation possible from whether it renews, depletes or leaves unchanged the capacities needed for future activity. Its broad philosophical account of autopoiesis is developed alongside more conditional claims about particular institutions and technologies.

Recent work on Rough Closure, The Mark in the Mesh and Toward a Mathematics of Rough Closure examines how definite operations coexist with consequential incompleteness. Ecological hyperknots describe differently organised processes that condition one another without becoming one undifferentiated system. Current formal work explores recursive stability, linkoids, boundaries and co-evolving conditions; its explanatory value requires careful specification and investigation.

A new discussion draft, Autopoiesis and the Mangle, engages Andrew Pickering’s account of practice, resistance and accommodation. It asks how the recurrence of an outcome differs from renewal of the capacity to produce it, and how persistence relates to novelty. The draft has been sent for discussion; the form of any joint work remains to be agreed.

2. Meaning mediation, education and mathematical proof

Meaning mediation examines how expressions acquire significance through interpretation and use. Semantic transduction follows what changes when an expression moves between settings—for example, when AI-generated text becomes a student explanation, an assessment judgement or an institutional record.

The September meaning-mediation book revision distinguishes technical output, understanding and warranted educational judgement. The recently published review with Christian Morgner and Erik Brezovec examines AI as meaning-mediating infrastructure, including inclusion and exclusion. Current work on Meaning Mediation in AI Assisted Mathematical Proving, developed from the Navier–Stokes discussion, asks how generated mathematical material becomes a claim that others can inspect, justify and accept.

3. Institutional programmes, authority and externalisation

Recursive Authorisation and the Reproduction of Institutional Programs investigates how selected outcomes and records renew a procedure’s authority. Continuing to produce recognised outputs, retaining procedural authority and achieving a substantive purpose are separate questions.

Compensatory work and the reproduction of organisational burdens examines how compensatory work can sustain apparent success while its burdens have little influence over the next allocation. It distinguishes a burden that is overlooked, one that is recognised without effective authority to change it, and one that is knowingly maintained. Each calls for a different account of power and repair.

New work with Gustavo Markier and Erik Brezovec asks how AI classifications acquire institutional authority through interfaces, documents and professional decisions. The developing article addresses programmatic closure and the politics of recognition. A small comparative pilot has been proposed; its settings and scope remain under discussion.

4. AEAI, constraint continuity and accountable control

The September AEAI architecture whitepaper centres on constraint continuity: whether a restriction retains its authorised scope and practical effect as a request becomes an interpretation, a delegated task, a tool call or a persistent record. The proposed architecture distinguishes interpretation, authorisation and execution. Recursive Purpose Mediation provides a way to reconsider an objective while requiring separate authority for expanded permissions.

The related article Control within and over collectives of artificial intelligence agents distinguishes local regulation among agents from people’s capacity to constrain their collective activity. It is the developed form of the earlier AI and human-extinction project. The AE Engine and AEAI remain a research and development programme: proposed specifications, constructed demonstrations and practical effectiveness require different evidence.

5. Political economy, interface debt and repair

The September Autopoietic Ecological Political Economy preprint develops AEPE through questions about finance, organisations, infrastructure, material resources and political authority. It introduces interface debt, translation power and repair capacity to investigate how burdens move between domains and why some losses become actionable while others do not.

Its contemporary case comparisons probe the plausibility of seven propositions, including visibility and repair, efficiency and fragility, inflation-policy specificity, inequality, and monetary–material sovereignty. The work develops and revises hypotheses rather than treating a small set of cases as definitive causal validation.

6. Learning, care, trust and collective life

Work with Gosia Marschall on affect, reflexivity and professional becoming develops an ecological account of teacher growth while retaining attention to teachers’ lived experience. Collaboration with Alastair Amos connects classroom uses of generative AI to dialogue, mathematical understanding and inclusion.

The revised urban-parks study with Andrew Charman and Heather Kruger examines differences in public accessibility in Delft, Cape Town. A chapter with Barry Gibson and Erik Brezovec explores trust in dental care. These projects retain their own questions, methods and responsibilities.

I-CIVIC: civic learning, trust and public agency

I-CIVIC gives the programme an international setting for examining democratic education. Steven’s Cambridge role focuses on training and knowledge exchange, connecting conceptual and methodological learning with the research undertaken across the partnership.

The wider intellectual question is how people acquire a meaningful capacity to participate. Information can be available while remaining difficult to interpret or challenge. A contribution can be acknowledged while having little effect on a decision. Civic education therefore raises questions about the relationships, resources and institutional responses through which participation becomes consequential.

Meaning mediation helps examine how claims and experiences change as they move through classrooms, platforms and public discussion. Work on authority and repair asks whether problems raised by participants can change subsequent action. These are proposed lines of inquiry, not established project findings.

The I-CIVIC page sets out project responsibilities, recent consortium activity and public sources. It distinguishes Steven’s contribution from the wider consortium’s work and from EdgeLab’s independent organisational status.

Current articles and collaborative writing

The following positions reflect the record checked on 18 September 2026:

• Control within and over collectives of artificial intelligence agents: submitted and acknowledged on 17 September.
• Compensatory work and the reproduction of organisational burdens: revised submission package prepared for Organization on 18 September 2026; not yet submitted.
• Why similar urban parks diverge: revised submission confirmed on 17 September.
• Generative AI, Technology, and the Ecology of Persistence: A Response to Tække: revised manuscript under review, based on the author’s current status record.
• Recursive Authorisation and the Reproduction of Institutional Programs: revised submission package prepared; submission not yet confirmed.
• Meaning Mediation in AI Assisted Mathematical Proving: submission package prepared; submission not yet confirmed.
• Affect, Reflexivity and Professional Becoming: advanced collaborative draft with Gosia Marschall.
• How AI classifications acquire institutional authority: development draft and small-pilot proposal circulated to Gustavo Markier and Erik Brezovec on 18 September.
• Autopoiesis and the Mangle: discussion draft sent to Andrew Pickering on 18 September.
• Luhmann, autopoietic ecology and the silent observation of trust in dental care: revised chapter for coauthor discussion.
• Afterword to Niklas Luhmann’s Political Sociology: production-stage work with Barry Gibson for the volume edited by Christian Morgner; final publication date awaits confirmation.

Four connected book projects

• Autopoietic Ecology: Rethinking Systems, Meaning, and Matter, with Erik Brezovec: a substantially revised manuscript following the 2025 public preprint; proposal submitted on 18 September.
• Generative AI and the Mediation of Meaning: Education, Society and Autopoietic Ecology: a revised manuscript developing understanding, judgement and participation in AI-mediated education; publisher submission package prepared.
• Living on the Edge: Why Our Systems Keep Failing Us—and What It Takes to Change Them: a general-reader account of institutional persistence and change; proposal and sample chapters submitted on 18 September.
• Living Under Pressure: Anxiety, Mental Health and Neurodiversity in Autopoietic Ecology: revised manuscript circulated for informal feedback on 16 September.

These are developing book projects. A proposal submission or reader-feedback request does not establish a publication contract.

How the programme will be investigated

The programme combines philosophical argument, engagement with published research, qualitative educational and social inquiry, comparative analysis and exploratory design. Three connected lines of investigation are being developed:

• Educational meaning and recognition: follow contributions through AI outputs, student explanations, assessment criteria and subsequent decisions, using documents, observation and participants’ accounts where appropriate.
• Institutional burdens and revision: compare how compensatory work, reports and recognised difficulties do—or do not—change later allocations and procedures.
• AI delegation and control: use bounded tasks and controlled comparisons to examine whether restrictions survive handoffs, changes of responsibility and interruption, and whether a person can intervene effectively.

These directions guide project development; they are not announcements of completed studies or confirmed partner commitments. Each inquiry needs a specified activity, boundary and timescale, plausible alternative explanations and evidence capable of challenging the proposed account.

Responsibility, participation and public engagement

Persistence is not sufficient evidence that an arrangement is desirable. Harmful practices can endure, and justified action can interrupt them. The programme connects explanation to explicit questions about bodily and psychological integrity, equal standing, participation, unequal authority and responsibility for remedy. Recording a concern is different from enabling those affected to alter what happens next.

EdgeLab’s Insights writing brings these questions into conversation with important global developments, across politics, the economy, climate, health, science, education, culture, technology and international affairs. News analysis uses the framework to formulate and assess interpretations while keeping reported facts, analytical claims and uncertainty distinct.

Research strands · People and collaborations · Publications and books · AE Engine and AEAI

We welcome shared questions, research settings and critical responses to the programme. Contact Steven Watson.