What Kind of Contribution This Might Be
The pattern
Across all domains searched -- behavioral health control engineering, adaptive automation, active inference HMI, personal informatics, SDT, exploitative technology design, rehabilitation robotics, cybernetics, and philosophical autonomy -- the specific integration was not found as a unified specification.
The individual components exist in different fields:
- Adaptive automation: closed-loop physiological monitoring
- Personal informatics: "observe, don't control"
- SDT: autonomy as a value
- Active inference: computational framework
- Schoeller: empowerment and virtual control
- JITAI lineage: control engineering applied to behavior
- Mayeroff: structural requirements for non-exploitative relationships
- Arnsten, Maier, Seligman: neuroscience of state-dependent capacity
- Rehabilitation robotics: "stabilize, don't replace" for motor skill
- Cybernetics: observer-included, model-based, autonomy-preserving
- Sen/Nussbaum: capability-over-functioning evaluative framework
- Frankfurt/Bratman: procedural autonomy account
- Grace/Inglis: DA gain control dynamics
Nobody has assembled these pieces because the fields don't talk to each other and the AI capability is new.
Strongest bridges
- Neuroscience of state-dependent capacity -> engineering constraints on human-AI design
- Capability approach -> real-time control objective
- Frankfurt/Bratman procedural autonomy -> measurable state variable
- Mayeroff caring -> testable system invariants
- Rehabilitation robotics "stabilize, don't replace" -> generalized to agency capacity
- Von Foerster's normative imperative -> implemented with modern neuroscience and AI
- Grace/MODAL DA gain dynamics -> connected to stress-capacity literature
Calibration
The contribution is synthesis, not discovery. The closest analogy is behavioral economics taking established psychology seriously as constraints on economic models. But that analogy is aspirational. Kahneman and Tversky had prospect theory, hundreds of experiments, and reproducible anomalies before the parallel was warranted. This work has a specification and proposed experiments. The structural parallel (established science from field A treated as binding constraints on field B) is the aspiration, not a claim of equivalent accomplishment.
Minimum viable contribution hierarchy
The contribution depends on what survives verification. From strongest to weakest:
If the unified specification is genuinely novel: The contribution is the specification itself -- the specific combination of control topology, objective function, neurobiological grounding, caring orientation, and design constraints that nobody has assembled. This is the strongest possible version.
If the specification is anticipated but not from this direction: The contribution is the independent derivation from neuroscience + control engineering + philosophy of agency, arriving at an architecture that someone in robotics or cybernetics specified from different premises. Convergent discovery from different starting points strengthens the case for the architecture.
If the specification is largely anticipated: The contribution shrinks to specific novel elements -- possibly the exploitative/supportive contrast formalized in control-theoretic terms, possibly the backfire regime as a formal prediction, possibly the objective function derivation from the definition of agency. Any of these could be a smaller, sharper document.
If everything is anticipated: The contribution is an integration document -- "here are the pieces that exist across fields that nobody has assembled." That is still valuable if the fields genuinely are not talking to each other, but it is a different document with different ambitions.
Which of these is the actual contribution remains partially unresolved. The remaining blind spots (shared autonomy, cybernetics, philosophical literature on autonomy) could move the answer. The document must be structured so that if the strongest version holds, it is clear; and if it does not, the weaker versions are still defensible without rewriting.
Completion criterion
The document is done when someone can read it, understand what is being argued, identify exactly where they disagree, and design an experiment to test the disagreement.