ARENOMICS INSTITUTE

Research

The Institute is research-led. This page sets out its founding research agenda, its working-paper programme, and the landscape evidence on which the Institute was founded.

Research agenda · working papers forthcoming

Research agenda

Six founding questions. Two are already active within The Synthetic Economy research programme and are hosted there, not owned here. Four are new to the Institute.

The price of mandate
The economic value of authority in a machine-labour economy. If permission and recognition can be priced, rented and captured at recognition gates, what does mandate cost, and who pays?
Verification asymmetry
The economics of verifying work produced by systems more capable than their verifiers. How much oversight to buy for a given residual risk, and with what statistical guarantee.
Composite liability
Accountability across agentic workflows that span several models, vendors and data sources — the case that existing designated-responsibility rules were not written for.
Delegation architecture
How authority propagates and attenuates through chains of humans, agents and sub-agents; when a delegated mandate should fail closed.
Synthetic enterprise design
The minimum governance structure for organisations with persistent autonomous agents, and the fleet-level economics of intervention as a finite resource.
Authority over autonomous cognitive production
The single-author position paper on the missing discipline: why the adjacent fields each decline part of it, and what the integrated field would have to answer.

Population-scale properties of delegated agency and the recognition boundary — who or what may be a legitimate economic actor — are active within The Synthetic Economy and remain that programme's work.

Working-paper programme

No Institute paper has yet been published. The intended first releases are a position paper on the missing discipline, and working papers on the price of mandate and on verification asymmetry. The Institute's own rule is that papers precede curricula: no programme requirement may cite an unpublished proposition.

Forthcoming. Dates are not set.

Source material

The Institute was founded on three internal documents completed in September 2026: a global landscape review of roughly 120 programmes, a white-space and future-professions study of 61 emerging role clusters and 17 adjacent disciplines, and a strategic concept document in which every claim is tagged as fact, inference, proposal or speculation. They are held as source material rather than published as Institute research. Their load-bearing findings are summarised below.

What the landscape review found

September 2026. Primary-source retrieval across the named territory; every finding traces to a page fetched during the review. The review was commissioned to argue against the Institute's founding claim, and partly succeeded.

What is not white space

Teaching people to build and deploy agents is a crowded market, from full master's degrees to a complete vendor tier. Agent governance is a named taught subject in professional certificates. Agents as economic actors are a research object at several universities and corporate institutes. The Technical University of Munich has taught students to conceptualise, design, implement and economically govern AI agents under its AGENTONOMICS framework since summer 2026. New master's degrees in AI governance, law and economics open at Oxford, Imperial and Birmingham from 2026–27. Four certification bodies already sell an AI-governance or AI-audit credential.

The claim that nobody teaches this is therefore false, and the Institute does not make it.

What is white space

No institution was found integrating agent design, mandate and delegation architecture, machine identity, verification and evidence, and synthetic-organisation governance into one professional formation pathway. Every pillar exists in a different building and a different discipline.

Mandate engineering and delegation architecture are the weakest pillars everywhere: no degree teaches delegation architecture as a named discipline. Machine identity is a standards field with no academic home. And a legally dated competence exists with no curriculum: EU AI Act Article 14 human-overseer duties begin 2 December 2027 and 2 August 2028, and no jurisdiction has specified who trains these people.

The review's own verdict was that the window in which a new institution can plausibly define this category is roughly 2026–2028.

Institutions doing adjacent work

Named in the review, without criticism: TUM (AGENTONOMICS); the University of Hong Kong's Centre for AI, Management and Organization; the Stanford Digital Economy Lab; MIT's Initiative on the Digital Economy, Center for Collective Intelligence and Project NANDA; Microsoft's AI Economy Institute; the UK's AI Economics Institute; Singapore's AI Verify Foundation, SMU Academy, NUS and NTU; Oxford, Imperial and Birmingham; the IAPP, IEEE, ISACA and ISO/IEC 42001 credential routes; and the Institute for AI Economics, an independent think tank. The Institute's position is to integrate, and where possible to seek mutual recognition, not to compete symmetrically.

What the review could not settle

Whether the gap is large enough to sustain an institute. Timing critics argue reliable agents are further away than the market assumes; industry analysts expect a large share of agentic projects to be cancelled; oversight itself is automating, with published intervention rates falling. The Institute's answer is that each of these increases rather than decreases the need for mandate, verification and revocation competence, and that the residual human function is smaller in headcount and higher in value. That answer is a hypothesis the Institute exists to test.

Trademark registers could not be queried programmatically during the review; naming findings were an open-web surface check and are being followed up separately.

Selected primary sources

The principal sources behind the public claims on this site, as fetched during the September 2026 review. Institutional and standards pages move; each link is to the page or record as it stood then.