Knowledge Object KO-004 · Self & Change

Research

Examine the evidence.

The evidence

What can actually be said about changing a learned pattern?

The sources below concern four separate questions: how associations are learned, how patterns are acquired through observation, how repetition in context produces habits, and how experience is associated with change in neural systems.

None of them establishes that a mind can be rewritten like software, and none of them equates human learning with machine learning.

Evidence map

Where the evidence is strong — and where it stops.

Associative learning

Established literature, reviewed

People learn relationships between events, stimuli, responses and outcomes; attention, memory and inference also shape what is learned.

Primary support

Shanks (2010), Annual Review of Psychology

Important nuance

Association is one mechanism among several. Human behaviour cannot be reduced to conditioning.

Observational learning

Established theoretical framework with extensive research

Behavioural patterns and information can be acquired by observing others, mediated by attention, retention, reproduction and motivation.

Primary support

Bandura (2001), Annual Review of Psychology

Important nuance

Observational learning is not simple imitation, and it does not explain every acquired pattern.

Habit formation

Annual Review synthesis

Responses repeated in recurring contexts can become efficient, cue-driven defaults.

Primary support

Wood & Rünger (2016), Annual Review of Psychology

Important nuance

Automaticity does not mean complete absence of agency, and habit strength depends heavily on context stability.

Experience-dependent plasticity

Experimental neuroscience review (largely animal models)

Experience and learning are associated with structural synaptic changes in the mammalian brain.

Primary support

Holtmaat & Svoboda (2009), Nature Reviews Neuroscience

Important nuance

Plasticity coexists with substantial stability. This does not support reprogramming a brain at will.

Intentional personality change

Systematic review of longitudinal studies

Intentional personality change is possible; interventions show small changes in the desired direction.

Primary support

Haehner, Wright & Bleidorn (2024)

Important nuance

Wanting to change is only weakly related to actual change; personality is not infinitely malleable.

Machine pattern learning

Technical review of machine-learning methods

Deep-learning systems discover statistical structure in training data by adjusting model parameters.

Primary support

LeCun, Bengio & Hinton (2015), Nature

Important nuance

This describes machine training procedures. It is not evidence about how humans learn, and the two are not equivalent.

Where this stands

Associative learning, observational learning, habit formation and experience-dependent plasticity are established research areas. The recoding metaphor, the human ↔ AI analogy as a source of meaning, and the inference that a learned pattern need not be treated as identity are creator interpretation, not findings.

People learn relationships between events, stimuli, responses and outcomes.
EvidenceEstablished
Behavioural patterns can be acquired through observing others, not only through direct consequences.
EvidenceEstablished
Responses repeated in recurring contexts can become efficient, cue-driven defaults.
EvidenceSupported
Experience and learning are associated with structural synaptic change in the mammalian brain.
EvidenceSupported
Intentional personality change is possible, and typically small.
EvidenceSupported
Deep-learning systems discover statistical structure in training data by adjusting parameters.
EvidenceEstablished
Human learning and machine learning are structurally analogous but mechanistically different.
InterpretationPhilosophical
If a pattern was learned, it does not have to be treated as identity.
PhilosophyPhilosophical
I recode myself as I come alive.
Artistic expressionUnverified

So what can we say?

Learned, changeable, and slower than the metaphor.

Patterns are acquired through experience, association, observation and repetition. They can change, through new learning, changed context and repeated experience. They can also persist, resurface and remain available long after they stop being useful.

Capacity for change is supported. A reset button is not.

The distinction

Two sentences that are not the same sentence.

Epistemic status: Core epistemic boundary of this object

“Humans and AI systems are both shaped by prior information.”

“Humans and AI systems learn in the same way.”

The first is a structural observation and the origin of the song's metaphor. The second is false, and nothing on this page should be read as supporting it.

Shaped by prior information≠Learning by the same mechanism

Human / AI

Same same, but very different.

Human

learns through lived experience
forms associations
can learn through observing other people
is influenced by emotion, motivation and social context
exists in a biological body and nervous system
has a developmental history

learns statistical patterns from training data
adjusts model parameters during training
does not have a human childhood
does not possess a human nervous system
does not acquire patterns through embodied human experience

SAME: prior information can influence future output or behaviour. DIFFERENT: the underlying mechanisms and the meaning of “learning” are profoundly different.

If both can learn patterns, what exactly do we mean when we say “learn”?

Boundaries