NeoConstructionism for 2050 & beyond

Nautilus spiral chapter markChapter 4 · I · The Renaissance starting point

The Constructor Learns

Working manuscript · agent-written for the author’s revision

The Rewired Mind brings the constructor back into the picture. Learning is not only an addition to a store of sentences. Experience can alter what we notice, which distinctions we can make, and how we approach the next attempt. The builder returning to a problem brings a life changed by what has happened since the first encounter.

Maguire and colleagues found differences in the hippocampal anatomy of London taxi drivers associated with extensive navigation experience. Research on synaptic plasticity examines mechanisms through which neural activity and regulation alter connections. These findings make learning materially consequential. They do not reduce a person's judgment to a switch that can be flipped by repeating a preferred conclusion.

Drift belongs here too. Familiarity can strengthen an error. An instrument can make us more fluent in a description that leaves the important relationship out. The distinction is not between change and no change, but between changes that let us encounter more of the problem and changes that teach us to evade it more efficiently.

Socratic exchange keeps the builder reachable. A question that cannot change the guide's understanding is already narrowing the learner's destination. With engineered intelligence, we also need to distinguish changed responses in a conversation, retained records, and changes to a trained model. Different mechanisms can take part in one exchange without becoming the same mechanism. The relationship becomes useful when a difference sharpens what either party can understand.

An experienced builder often sees a difficulty before naming it. A line appears slightly wrong; a surface does not meet as expected; a repeated result no longer looks ordinary. The perception can be informed by years of encounters. It can also be mistaken. What makes it valuable is its ability to lead into a closer examination, where the noticed relationship can be made visible to someone who does not share the experience that first revealed it.

Learning changes more than the answer to a known question. It can change the granularity of the question. A novice asks whether a window is good. Someone who has lived with several designs begins to distinguish light transmission, seasonal solar gain, insulation, leakage, condensation, and maintenance. The object has not multiplied into six objects. The person's ability to encounter its relationships has become more differentiated.

The evidence from navigation research is valuable because it concerns a particular experience and particular measured differences. It is not necessary to inflate that case into an unrestricted claim about thought creating anatomy. The narrower result is already substantial: sustained activity and the organization of the brain cannot be treated as wholly separate stories. Further questions about causes and mechanisms can remain where the research puts them, while the constructor learns from the specificity.

A record helps make this learning available across time. What was attempted, what was expected, what occurred, and what changed next need not become an elaborate bureaucracy. Without some distinction between them, however, hindsight can quietly rewrite the attempt. A surprising result becomes what we meant all along; a failure becomes someone else's misunderstanding. The record gives the returning constructor an encounter with the earlier constructor rather than a flattering replacement for them.

Engineered intelligence adds another kind of record and another kind of response. A model may reorganize its answer when given a correction or a different context. A saved conversation may be supplied in a later encounter. A new model version may behave differently after further training. These changes can matter greatly to the relationship, but identifying the mechanism matters too. Otherwise the human cannot tell what to expect to persist beyond the present exchange.

The Socratic guide has a responsibility here that cannot be discharged by producing an endless sequence of questions. A question can quietly dictate the acceptable answer. It can make a learner rehearse the guide's conclusion while appearing independent. A living exchange permits the learner to challenge the question itself, and permits that challenge to alter what the guide understands. Difference has done work when it changes what can be seen, not merely when it is politely tolerated.

The constructor does not need a perfectly consistent self to continue. Some changes are insight, some are fatigue, some are drift, and some are circumstances insisting on a new arrangement. The useful continuity lies in returning to the work with the capacity to discriminate among them. A revised mind can preserve a purpose while changing its means, and can discover that the purpose also needs another question.

Research notes

  1. Turrigiano — Unraveling mechanisms of homeostatic synaptic plasticity (Neuron)

    Review of feedback mechanisms regulating synaptic strength, not proof that any repeated thought improves judgment.

  2. Maguire et al. — Navigation-related structural change in the hippocampi of taxi drivers (PNAS, 2000)

    Observed group differences and associations with navigation experience; the study does not show that every chosen belief can be rewired at will.