The 2029 framework

Accessibility as multi-agent communities of practice with formal equilibrium dynamics. The next research step explicitly named in the 2013 thesis conclusion, paused because the implementation tools didn’t exist, and resumable now that practical agentic AI has arrived.

What the framework claims

Accessibility is not a binary pass/fail and not a static property of either system or user. It is an emergent dynamic equilibrium in a contextual space of competition between five factors:

  • environmental factors,
  • technical constraints,
  • user capability,
  • user preference, and
  • available UI resources and modalities (the limited resource).

The framework requires agency for each factor — each must be able to advocate, negotiate, and respond.

The published seed (2006)

The framework was first named in print in the 2006 short paper:

“If the role of each agents involved in interface construction is definable then a definition of (at least) intrinsic accessibility using formal methods and game theory should be possible.”

The 2013 thesis conclusion named the same step, explicitly:

“That competition, I would suggest, would be best considered through the use of game theory ... whether the mathematical formalism my research requires will be found in models of autonomous agents.”

In 2006, the theory was understood; the compute was not. In 2013, the formalism was clearer; autonomous-agent implementations remained academic toys. With practical agentic AI now available, the implementation tools have caught up. This is the line — the 2029 project is the resumption of explicitly-named research, not a hobby waiting for retirement.

Why competition alone is the wrong frame

“Game theory and autonomous agents” alone undersells the framing. The earlier Communities of Practice chapter had the deeper move: agents are not just competitors, they are members of a community of practice with collaboration, role-playing, shared enterprise, and memetic learning.

Today’s multi-agent vocabulary (auctions, negotiations, marketplaces, multi-agent reinforcement learning) is competition-shaped. The CoP framing is collaboration-and-learning-shaped — agents tuning their relations, learning collectively, holding shared identity, migrating between roles. Same agents; richer model. The contemporary agentic-AI literature is mostly the first half. The second half — communities of practice with multi-agent dynamics — is materially open.

Whether the game is cooperative or non-cooperative is the wrong question. The actual question is how do agents in a community of practice negotiate competition for resources while sustaining the shared enterprise that makes the cooperation valuable in the first place?

The roadmap

The 2029 project is the resumption of named research: applying the named-in-thesis-conclusion theoretical framework to accessibility, using agentic AI as the implementation substrate that wasn’t available in 2013, and explicitly carrying both the game-theoretic competitive dimension and the CoP collaborative-and-learning dimension.

This page is the public roadmap. Specific framings, prior-art surveys, and any associated publications will be added as that work resumes.

Reading on