Adversarial Systems Research

The Adversarial Systems & Complexity Research Initiative studies complex multi-agent behaviour and its adaptive mechanisms. Dissensus proposes that the pursuit of non-stable equilibria functions as the underlying stabilising condition. That lens drives computational, empirical, and theoretical work on risk, coordination, and adaptation.

Focus
Adversarial systems
Publications
0 papers
Status
Active research
arXiv ASRI 2602.03874·Replicator-Optimization Mechanism 2601.06363·Axiom of Consent 2601.06692·Semantic-First Vision 2603.04419
Under review Digital Finance · AI and Ethics · Ethics and Information Technology · Computational Economics · Financial Innovation
The archive — drag to explore

dis·sen·sus /dɪˈsɛnsəs/, from Latin dissentire, to disagree. In political philosophy (Rancière, Mouffe), dissensus names the productive disagreement that remains when consensus is impossible or undesirable. The lab is named for it. Not dysfunction, function.

01 · Who we are

An independent lab, built around one programme

Dissensus is a UK research lab investigating adversarial systems — where competing interests generate structural conflict. Friction dynamics is the through-line, from markets and institutions to cognition and AI.

Dissensus is built around a single ongoing programme. ASCRI investigates sufficiently complex dynamic networks of agents where equilibria cannot be reached under formalised friction constraints. Such systems function only through the continual reduction of friction: they are adversarial against themselves, yet persist through dissensus. Every paper in the archive probes one face of this programme — consent mechanics, market microstructure, machine morality, or the process-philosophical foundations underneath.

Active · hosted at systems.ac →
0
Publications
0
Research domains
0
Peer-review venues
0+
Machine-checked proofs
02 · The problem

Before strategy, there is delegation

Game theory presupposes participation. It does not account for how agents enter strategic interaction, why payoff structures take the forms they do, or what happens when participation is refused. It does not interrogate who defines the rules, whose interests they serve, or whether the game itself is legitimate. We study the layer underneath. Before strategy, there is delegation — and delegation produces friction.

F = σ (1 + ε) 1 + α
Friction, decomposed — it rises with stake and entropy, and falls with alignment. A single functional, three measurable primitives.
α
Alignment
How closely the receiver's interests track the delegator's. The one term that reduces friction.
σ
Stake
How much is at risk in the delegation. Scale amplifies every misalignment beneath it.
ε
Entropy
How much the receiver misreads the delegator's preferences. Uncertainty is never free.
03 · Contribute

Three ways into the lab

Dissensus is small, independent, and open at the edges. The research runs regardless, but it runs faster and argues better with more people in the room.

I
Fund
Money buys research time and nothing else. Funds go directly to research and research collaboration — compute, data access, publication costs — and we will keep a running public account of where they go. What that supports right now is listed under active projects. Funding never buys editorial reach.
II
Propose
Three research directions are currently accepting proposals: adversarial systems & AI governance, systemic risk measurement, and philosophy of AI. Bring a domain, a dataset, or a critique — strong counterarguments get published with full attribution. The proposal form takes five minutes.
III
Join
Research collaborators, research assistants, co-authors. There are no salaried positions, and it is honest to say that first. What the lab offers is your name on the work you help produce, and time from someone who will actually read your drafts. Students and early-career researchers are welcome; affiliation is optional, rigour is not.
04 · Proposal

Propose a project

Complete the form below to submit a proposal. We respond to all substantive enquiries within five business days.

05 · Elsewhere