Hume vs. Kant on the Human Mind - What Modern Psychology and Neuroscience Support

Verdict

There is no clean scientific winner because Hume and Kant made different kinds of claims, many of them philosophical rather than experimentally testable. Still, the pattern is striking:

• Motivation, emotion, habit, and much ordinary moral judgment: the evidence is more Humean.
• Perception, early cognitive structure, and the mind’s active organization of experience: the evidence is more Kant-like.
• Moral cognition as a whole: contemporary research points to an interaction among affect, intuitive valuation, social learning, intention-reading, rule use, and deliberation—a Hume–Kant synthesis rather than either philosopher in isolation.

The Righteous Mind, Chapter 6 — Haidt’s Humean verdict is descriptive and source-era

Haidt favors Hume’s observational and sentimentalist program over attempts to derive a complete moral psychology from pure reason. His source-era Figure 6.1 adapts Baron-Cohen’s empathizing/systemizing dimensions and places Kant in a high-systemizing, low-empathizing region to suggest why a single universal formula may underrepresent lived motives and relationships.

That image cannot carry a scientific verdict on Kant. Haidt’s placement is interpretive, the biographical basis for retrospective diagnosis is unclear, and the framework is not a current diagnostic map or evidence about autistic morality. Even if Hume better describes how ordinary judgment works, that would not refute Kant’s normative arguments or his claims about the conditions of rational obligation. Hume’s observational program, PDF pp. 139–140 (file not included in this release) · Figure 6.1 and Kant, PDF pp. 142–147 (file not included in this release) · Chapter 6 source audit

Evidence boundary

What the science can—and cannot—establish

The labels Humean and Kant-like mark structural affinities between empirical findings and philosophical pictures. They do not show that an experiment has proven Hume’s moral philosophy, Kant’s transcendental idealism, the categorical imperative, or any metaphysical thesis about causation or the self. The numerical scores below are interpretive heuristics, not results of a meta-analysis.

The fairest comparison separates at least four questions:

1. What moves people to act?
2. How do people actually make moral judgments?
3. How does the mind turn sensory input into organized experience?
4. What follows normatively—what people ought to do—from any psychological finding?

The fourth cannot simply be read off the first three. Descriptive psychology can constrain a moral theory’s psychology, but it does not by itself settle moral truth. See Hume Is-Ought Problem.

The dispute in one map

1. Emotion and motivation — the strongest Humean result

Hume’s slogan is often quoted without its argument:

“Reason is, and ought only to be the slave of the passions.” — A Treatise of Human Nature, 2.3.3.4

He does not mean that reasoning is useless or that one should impulsively obey every emotion. His central claim is that reason, considered only as the representation of truth and falsity, does not generate an ultimate end. Reason can reveal facts, compare means, calculate consequences, and correct mistaken beliefs. To produce action, however, cognition must connect with a desire, aversion, sentiment, commitment, or other conative state.

This is the issue treated more fully in Hume, Stoicism, and Whether Reason Can Motivate Moral Action.

The lesion evidence

In the classic Iowa Gambling Task study, Antoine Bechara and colleagues examined decision-making associated with damage to the ventromedial prefrontal cortex. Patients could retain substantial intellectual ability and explicitly discuss choices, yet repeatedly selected disadvantageous options. Healthy participants developed anticipatory bodily responses before choosing from risky decks; the affected patients did not show the same useful anticipatory pattern.¹

The broad lesson fits Hume better than a model of practical agency as detached calculation:

representation + valuation + action-selection, not reason alone → action

This also fits Antonio Damasio’s later “somatic marker” interpretation: affective signals help rank options when exhaustive calculation would be too slow or underdeterminate. See Descartes' Error - Emotion, Reason, and the Human Brain.

Important qualification

The strongest popular interpretation of the original task—participants’ bodies “knew” which decks were bad before participants possessed relevant conscious knowledge—has been disputed by later work about awareness, measurement, and task strategy. The safer conclusion is not that unconscious emotion mysteriously solves decisions before cognition begins. It is that effective decision-making depends on integrated valuation signals and cannot be reduced to intact abstract intelligence.

Why this does not eliminate reason

The result undermines motivational intellectualism, not reasoning. Deliberation can:

• alter beliefs about which acts serve an end;
• make distant consequences salient;
• reframe what a person values;
• inhibit a locally attractive impulse;
• construct plans, rules, and implementation intentions;
• educate emotion through repeated attention and practice.

Modern work therefore weakens the picture of reason and passion as two independent faculties fighting for control. Cognition and affect are deeply interdependent processes.

2. Moral judgment — Hume wins the opening round

The emotion-centered turn in moral psychology was hostile to a simple model in which people first derive a principle through conscious reasoning and then apply it dispassionately.

Joshua Greene and colleagues found different patterns of neural engagement in “personal” and “impersonal” moral dilemmas. Personal-harm dilemmas recruited regions associated with emotional and social processing more strongly, while some difficult conflicts produced longer response times and greater engagement of deliberative control processes.²

Likewise, experimentally increasing serotonin changed judgments and behavior concerning personally harming another person, especially where aversion to harm was central.³ Such findings make it difficult to treat affect as merely noise added after an otherwise complete moral calculation.

This supports a restrained Humean thesis:

Sentiment is often constitutive of ordinary moral evaluation, not merely an irrational obstacle to it.

It does not show that every emotion is morally reliable. Disgust, anger, empathy, and loyalty can each track morally relevant features while also producing bias, parochialism, or manipulation.

3. Psychology did not show that Kant was simply wrong

The first wave of “intuition first” research was sometimes summarized as though conscious reasoning were always post-hoc confabulation. That is too strong. People also represent:

• what an agent intended;
• what the agent believed;
• whether an outcome was foreseeable;
• alternative actions and counterfactual outcomes;
• fairness and reciprocal expectations;
• rules, precedents, and institutional roles;
• conflicts between rights, welfare, loyalty, and desert.

Liane Young and colleagues found that moral judgment recruits theory-of-mind processes that represent another person’s beliefs and intentions.⁴ Two outwardly identical harms can therefore receive different judgments because one was intentional and the other accidental. That is structured cognition, not an undifferentiated feeling of disapproval.

Kant should not, however, be reduced to the empirical prediction that people consciously perform syllogisms before acting. His central moral claim is normative: a rational will ought to act on principles it could will universally and to treat persons as ends. Showing that ordinary judgment is emotional would not by itself refute that standard.

4. A genuinely Kant-like capacity: universalization

Sydney Levine and colleagues tested whether people evaluate a tempting act partly by asking what would happen if everyone in the relevant situation acted on the same rationale. Their experiments found that a formal model of universalization helped explain judgments in threshold-dependent social dilemmas.⁵

The resemblance to Kant is obvious:

Can I rationally endorse everyone acting on this maxim?

But the resemblance has limits. The experiments concern an empirically observed style of reasoning, often sensitive to predicted aggregate consequences. Kant’s categorical imperative is a normative test grounded in autonomy and rational consistency. A psychological tendency to universalize is therefore Kant-like, not experimental proof of Kantian ethics.

5. The interactive model of moral cognition

The evidence fits neither “emotion makes the judgment and reason merely invents excuses” nor “pure reason computes the answer and emotion only interferes.” A better map is:

flowchart TD
    S["Situation"] --> A["Affective appraisal"]
    S --> M["Mental-state model"]
    S --> R["Rules and learned norms"]
    A --> I["Initial intuition"]
    M --> I
    R --> I
    I --> D["Deliberation and counterfactuals"]
    D --> J["Moral judgment"]
    I --> J
    J --> L["Social feedback, learning, and habit"]
    L --> A
    L --> R

A 2022 study strengthens this hybrid picture: intuitive judgment itself can make structured trade-offs among competing moral values.⁶ “Intuitive” therefore need not mean cognitively simple, and “reasoned” need not mean affect-free.

The Humean and Kantian components can be represented as different emphases inside one recurrent system:

• Humean: motivational force, sympathy, aversion, habit, learned association, rapid valuation.
• Kant-like: rule representation, intention-sensitive judgment, consistency checks, universalization, reflective self-government.

6. General cognition — Kant’s major comeback

A crude empiricist model treats the mind as a passive container:

sensations arrive → associations accumulate → concepts appear

Kant’s alternative was that sensations become coherent experience only through the activity of the mind. Space, time, causality, objecthood, and unity were not extra sensory objects; they were conditions under which an ordered world could be experienced.

Modern cognitive science does not establish Kant’s specific table of categories or transcendental idealism. It does, however, overwhelmingly reject the mind-as-camera picture. Perception depends on active inference, selective attention, prior learning, task demands, and recurrent top-down processing.

That makes modern cognition Kant-like in architecture, even where its mechanisms are naturalistic, probabilistic, learned, and neurally implemented in ways Kant could not have anticipated.

7. Infant cognition — early structure, carefully interpreted

Developmental research reports surprisingly early competencies involving objects, agents, quantities, probabilities, speech patterns, and causal structure.

• Three-month-old infants represented observed actions in terms of cause, cost, and goal, although they could not perform those actions themselves.⁷
• Newborns’ brains detected certain structural regularities in speech, revealing very early sensitivity to abstract patterns.⁸
• Twelve-month-olds formed probabilistic expectations about future object configurations in a visual scene.⁹

These findings make an unrestricted “everything is acquired by accumulating simple associations” account implausible. The infant is not a blank recorder.

Early does not automatically mean innate

An ability measured in a newborn may reflect prenatal learning, rapid domain-general learning, built-in biases, evolved domain-specific structure, or some interaction among them. Looking-time and neural-response paradigms also require careful interpretation. Early competence supports the claim that cognition brings structure to experience; it does not by itself identify that structure with Kant’s categories.

8. Perception is constructive, but not purely a priori

An infant-neuroscience study found learning-induced top-down influence from higher-level predictive information onto lower-level sensory processing.¹⁰ More generally, contemporary perception research treats expectations and prior information as factors that shape how signals are interpreted.

This is Kant-friendly because the experienced world is not a raw copy of sensory input. Yet it is also Hume-friendly because the priors and predictive models are extensively updated through exposure, association, and habit.

The synthesis is therefore:

The mind actively organizes experience, and experience continuously reorganizes the mind.

Kant is strongest against passive receptivity; Hume is strongest against a picture in which learning and habituation are secondary.

9. Causality — each philosopher captures a different problem

Hume’s problem

When we observe one event followed by another, we see succession and regularity, not an additional sensory item called necessary connection. Repetition produces an expectation that B will follow A. Hume’s epistemic warning remains powerful: statistical and causal models describe dependencies under specified conditions; they do not reveal metaphysical necessity by direct inspection.

Kant’s reply

Kant thought that mere association could not explain how we experience an objective temporal order in the first place. Causal organization is part of what makes a succession intelligible as an event in one shared world rather than a private sequence of impressions.

Early causal cognition lends psychological support to the idea that humans approach experience with powerful organizing capacities. But those capacities may be probabilistic, revisable, and partly learned rather than fixed a priori categories in Kant’s technical sense.

10. The self — bundle and unity

Hume reported that introspection encounters particular perceptions—sensations, emotions, memories, and thoughts—but no separately presented, simple ego. Modern neuroscience likewise finds no single place where a metaphysical self sits. Bodily ownership, agency, autobiographical memory, narrative continuity, perspective, and social identity depend on distinguishable and distributed processes.

This is compatible with a constructed or bundled self. Disturbances and experimental manipulations of agency or body ownership further show that components of selfhood can dissociate.

Kant anticipated the complementary difficulty. His “transcendental unity of apperception” is not a little object hidden in the brain. It is the requirement that different representations be combinable within one point of view—capable, in principle, of accompanying the thought “I think.” Cognitive neuroscience does study the binding, access, memory, and integration mechanisms necessary for coherent conscious perspective.

So neither simple victory follows:

• Hume: no separately observable metaphysical ego has been discovered.
• Kant: organized cognition nevertheless requires functional unity across contents and time.

Science can investigate the functional organization of selfhood. It cannot infer merely from distributed processing that no metaphysical subject exists, nor from integration that Kant’s transcendental subject has been proven.

11. Hume’s scorecard

12. Kant’s scorecard

13. A heuristic verdict—not a scientific score

If forced to compress the comparison into numbers:

These numbers are mnemonic summaries of the argument, not measured effect sizes. Altering which claims count—or distinguishing Kant’s normative theory from a caricature of rationalist psychology—would alter the score.

14. Relation to Robertson’s The Enlightenment

This bears directly on The Enlightenment - Ritchie Robertson. A historical story in which Stoic or Kantian rational self-command is simply superseded by Humean and Smithian sympathy would be too linear as a statement about the human mind.

Modern psychology supports the importance of sentiment, sympathy, and affective valuation. But it also supports rule use, self-regulation, intention representation, universalization, and active cognitive structuring. The empirically plausible picture is not:

rational control or sympathy

but:

emotion + sympathy + reason + learned norms + self-command + structured cognition

That is also why the contrast with Stoicism needs care. Stoic knowledge is meant to be an embodied and practiced transformation of judgment, not mere possession of a detached proposition. See Hume, Stoicism, and Whether Reason Can Motivate Moral Action.

15. Best synthesis

The contemporary position

Hume was right that human beings are affective, motivational, social, and habit-forming animals, and that reasoning normally operates inside a system of concerns and valuations.

Kant was right that experience is not a passive stream of impressions. The mind actively organizes information, supports rule-governed reflection, and integrates representations into a coherent point of view.

Both become misleading when isolated. Emotion can contain sophisticated structure; reasoning can transform emotion and motivation; innate constraints and learning interact; and moral cognition is simultaneously affective, social, predictive, and deliberative.

The most defensible verdict is therefore domain-specific pluralism. Hume has aged better as a theorist of motivational force and sentiment. Kant has aged better as a theorist of constructive cognition. Modern science supplies mechanisms neither possessed and does not settle their deepest normative or metaphysical disputes.

Confidence map

Sources

Primary empirical studies

1. Bechara, Damasio, Tranel, and Damasio, “Deciding Advantageously Before Knowing the Advantageous Strategy,” Science 275 (1997): 1293–1295. PubMed.¹
2. Greene, Sommerville, Nystrom, Darley, and Cohen, “An fMRI Investigation of Emotional Engagement in Moral Judgment,” Science 293 (2001): 2105–2108. PubMed.²
3. Crockett, Clark, Hauser, and Robbins, “Serotonin Selectively Influences Moral Judgment and Behavior Through Effects on Harm Aversion,” PNAS 107 (2010): 17433–17438. PNAS.³
4. Young, Cushman, Hauser, and Saxe, “The Neural Basis of the Interaction Between Theory of Mind and Moral Judgment,” PNAS 104 (2007): 8235–8240. PNAS.⁴
5. Levine, Kleiman-Weiner, Schulz, Tenenbaum, and Cushman, “The Logic of Universalization Guides Moral Judgment,” PNAS 117 (2020): 26158–26169. PNAS.⁵
6. Levine, Chater, Tenenbaum, and Cushman, “A Moral Trade-off System Produces Intuitive Judgments That Are Rational and Coherent,” PNAS 119 (2022): e2214005119. PNAS.⁶
7. Liu, Brooks, and Spelke, “Origins of the Concepts Cause, Cost, and Goal in Prereaching Infants,” PNAS 116 (2019): 17747–17752. PNAS.⁷
8. Gervain, Macagno, Cogoi, Peña, and Mehler, “The Neonate Brain Detects Speech Structure,” PNAS 105 (2008): 14222–14227. PNAS.⁸
9. Téglás et al., “Pure Reasoning in 12-Month-Old Infants as Probabilistic Inference,” Science 332 (2011): 1054–1059. Science.⁹
10. Emberson, Richards, and Aslin, “Top-Down Modulation in the Infant Brain: Learning-Induced Expectations Rapidly Affect the Sensory Cortex at 6 Months,” PNAS 112 (2015): 9585–9590. PNAS.¹⁰

Philosophical anchors

• David Hume, A Treatise of Human Nature, especially 2.3.3 and Book III.
• Immanuel Kant, Critique of Pure Reason and Groundwork of the Metaphysics of Morals.
• Hume, Stoicism, and Whether Reason Can Motivate Moral Action.
• Descartes' Error - Emotion, Reason, and the Human Brain.

Footnotes

1. This study is foundational, but later debate weakens the strong “body knows before conscious cognition” interpretation. The integrated role of valuation in advantageous decision-making is the more defensible takeaway. ↩ ↩²

2. The contrast among dilemma types should not be treated as a literal brain map of “emotion versus reason”; the tasks recruit overlapping distributed processes. ↩ ↩²

3. A pharmacological effect on harm aversion shows causal sensitivity of judgment to serotonin-related processing, not that morality reduces to one neurotransmitter. ↩ ↩²

4. Intention-sensitive judgment requires structured representations but does not, by itself, favor Kant’s normative conclusions. ↩ ↩²

5. Universalization-like reasoning can be consequentially sensitive and is not identical to the categorical imperative. ↩ ↩²

6. Structured intuitive trade-offs challenge a strict equation of intuition with simplicity or irrationality. ↩ ↩²

7. The findings concern early representations of observed action; they do not demonstrate the entirety of Kant’s causal category. ↩ ↩²

8. Neural discrimination of speech structure is evidence of early pattern sensitivity, not a general proof of innate grammar. ↩ ↩²

9. Looking-time measures support probabilistic expectations under the experimental model; they do not show that infants consciously calculate numerical probabilities. ↩ ↩²

10. Learning-induced top-down modulation supports constructive perception while also demonstrating the importance of experience. ↩ ↩²