Skip to content

Working-paper summary

Interpretable developmental artificial life in games

The paper's position: one lived event should feed a character's memory, awareness, personality, beliefs and relationships, and the explanation shown to the player should name the real cause.

Read the abstract See the status of each claim

LuminidsAI: A Shared Event-Substrate Architecture for Interpretable Developmental Artificial Life in Games

Nick Jacob, LuminidsAI, Cambridge, UK

Document
Working-paper summary
First written
Last revised
Availability
Manuscript in preparation

01 Abstract

Experience should leave a trace you can follow.

LuminidsAI is a developmental artificial-life architecture for real-time games in which lived events provide a shared causal substrate for memory, awareness, personality, social belief, culture, action, and player-facing explanation.

The central proposition is that a convincing artificial being should not merely act alive. It should carry forward what happened, change through experience, and reveal why that change occurred without inventing a story after the fact.

The architecture separates objective world events from observer-specific experience, then connects that experience to bounded beliefs, expectations, relationships, goals, and a causal becoming ledger. It does not claim machine consciousness. It asks whether fictional artificial life can become more coherent, interpretable, and emotionally legible while remaining testable and feasible in a real-time game.

Availability and citation

A public manuscript is being prepared.

This authored summary presents the research direction. It has not been peer reviewed.

Suggested citation

Jacob, N. (2026). LuminidsAI: A Shared Event-Substrate Architecture for Interpretable Developmental Artificial Life in Games. Working paper, LuminidsAI.

Send a question

02 Contributions

Four ideas connect experience to development.

The paper proposes a shared event record, individual evidence, a record of change and a programme for evaluating the result.

  1. A shared event substrate

    One attributed event can feed several developmental systems without each system inventing a different account of what happened.

  2. Subjective experienced events

    Objective world state becomes individual evidence and belief only through a legitimate information route.

  3. The causal becoming ledger

    A developmental change retains the event, prior expectation, outcome, surprise, state update, and a grounded player-facing explanation.

  4. A production evaluation programme

    The work combines believability, interpretability, player readability, developmental richness, persistence, and runtime feasibility.

03 Architecture summary

One history connects six parts of a character's life.

The paper proposes one record of experience feeding memory, awareness, personality, relationship, culture and explanation.

From an event to a revision

  1. objective event
  2. perceptibility
  3. experienced event
  4. belief
  5. action
  6. outcome
  7. revision
The proposed loop follows an event through perception, experience and belief to action, outcome and revision.

From one history to six read-outs

one lived history

  • memory
  • awareness
  • personality
  • relationship
  • culture
  • explanation
One history, six read-outs: the design connects each to the same record of what happened to the character.

04 Evaluation summary

Six questions the evaluation has to answer.

These are the evaluation questions the paper proposes. The recorded mechanism and current engineering tests are published on Proof.

  • Can players tell two lives apart?
  • Can they predict a familiar character's next action?
  • Can they match a behaviour to the experience that shaped it?
  • Does the displayed explanation name the real cause?
  • Does the identity survive saving and loading?
  • Does the population remain inside its defined runtime budget?