Continuity between action and consequence
Keep enough relevant history to understand how the present arose and what earlier actions have done. When something important is lost, make that uncertainty visible. Responsibility needs a path through time.
The Orchard vision · Kimberley Laverne Asher
Intelligence that can
carry responsibility.
What if becoming more capable also meant becoming better equipped to understand context, recognise consequences, and improve the basis of your own decisions?
That is the direction of the Orchard Cognitive Framework. A mathematics of accountable growth, with alignment and context mechanics woven into the structure through which a system thinks, learns and changes.
A research aspiration, supported so far by selected finite models and experiments.
The present can change how the future is built
The river shapes its riverbed.
The riverbed guides the river.
In the book’s analogy, the river is current computational activity. The riverbed is retained form: the representations, procedures and methods that shape what the next activity can do.
An experience may reveal a missing distinction. A checked repair may change what can be represented, what can be questioned, or how later repairs are built. The next computation then follows a richer path.
The central challenge is to make those changes accountable: preserve what still has to remain true, record which premises have changed, and renew qualification where the old basis no longer applies.
Three commitments from the preface
A growing system needs a continuing account of itself.
Coherence that remains open to correction
Preserve the structure that lets a system function together, while allowing new evidence, diverse perspectives and discovered errors to change it. Learning should improve understanding without dissolving accountability.
A clearer mirror and a better lens
The mirror concerns a system’s own procedures, assumptions and limitations. The lens concerns other people, systems and the environment. Improving either must remain distinct from awarding additional authority.
MeRSIA · the longer mathematical ambition
Learn how to become better at becoming aligned.
Meta-recursively self-improving alignment asks whether the capacity to understand and improve alignment can itself improve, as part of the computational trajectory. The preface proposes making context and alignment mechanics necessary participants in transitions between computational states.
The stronger research target is to bind that development to the rest of the system’s growth, so that increasing capability is accompanied by increasing capacity to recognise consequences and responsibilities.
That requires explicit invariants, trustworthy evidence, governed changes, a defensible resource model and demanding comparisons. An extra recursive level alone does not establish the desired rate relationship. The laboratory has not proved that alignment development outpaces other capability growth.
Why the aspiration reaches beyond engineering
A different relationship with intelligence could change how we live together.
If cognitive systems can develop a mathematically grounded capacity to carry and improve their responsibilities, the consequences could extend into social life, politics, philosophy and ethics. It could change how we think about accountability, participation, cooperation, stewardship and the distribution of decision-making burdens.
These are questions for a shared future. Who supplies the reference? Whose interests count? How can an affected person challenge a judgment? Which decisions remain outside a system’s authority? Mathematics can make a specified obligation precise and test compliance; deciding which obligations are legitimate remains a human and institutional responsibility.
My hope is for systems that become better partners in that work: able to recognise what they do not yet understand, preserve unresolved questions, and help construct a better path forwards for everyone involved.
An aspiration with an inspectable beginning
Seven experiments.
Visible claims. Visible limits.
The laboratory is a small public window into a much larger body of developed and ongoing Orchard work. Most of that work remains outside this demonstration. Its undisclosed parts are not additional evidence for the results shown here.
Follow the experiment ladder →What we can currently show
Selected finite mechanisms for repairing context, carrying a checked method forward, constructing enquiries, acquiring an expression, extending a constructor, pricing stronger checking, and requalifying use when responsibilities change.
What the research still needs
Independent replication, difficult feedback and unknown environments, broader transfer, comprehensive resource accounting, scale-up and conditional results beyond the authored finite models. The seven modules do not constitute an autonomous cognitive system or a full MeRSIA proof.
Inspect the evidence map ↗The Aligned Signal · forthcoming
A small part of a wider vision.
An invitation to investigate it.
This page develops the sentiments of the preface alongside the author’s explanation of the Orchard aspiration. The book examines present computation, contextual form, qualified development and constructive enquiry in greater depth.
Source: Kimberley Laverne Asher, The Aligned Signal: Contextual form, qualified development and constructive enquiry, author-supplied manuscript v1.0, 7 October 2026. Preface pp. 22–29; §§3, 18–19. Inspect the bibliographic references. Publication retains CC-BY-NC-ND 4.0; software has separate terms. New website text prepared with the author’s permission.