Leveraging psychedelic neuroscience to boost human creativity using artificial intelligence

The paper addresses whether and how artificial intelligence can augment human creativity by emulating cognitive and neural mechanisms implicated in psychedelic states. Framed against societal challenges and the notion that people operate within entrenched "reality tunnels," the author posits that both psychedelics and AI can induce productive cognitive destabilization that enables novel insights. The purpose is to synthesize psychedelic neuroscience and psychology (e.g., Default Mode Network disruption, reduced latent inhibition, increased divergent thinking, enhanced implicit learning) and map them to AI design strategies that support creativity. The paper argues for AI that works with human psychology—adaptive to context, personality, and goals—rather than around it, and outlines risks and safeguards necessary for responsible implementation.

Section 2 reviews psychedelic psychology and neuroscience: classic psychedelics (e.g., LSD, psilocybin, DMT) act primarily via serotonergic mechanisms and consistently modulate the Default Mode Network (DMN), increasing brain-wide connectivity and reducing self-referential processing. This supports more entropic, flexible cognition and can enable insight, with context (set and setting) strongly moderating outcomes. Section 3 reviews psychological foundations of creativity, emphasizing three constructs: (i) latent inhibition—filtering of previously irrelevant stimuli—where lower levels can broaden associative access and support novelty; (ii) divergent versus convergent thinking—idea generation versus evaluation/selection; and (iii) implicit learning—unconscious pattern acquisition that fuels intuitive insight. Evidence links psychedelics to reductions in latent inhibition, increases in divergent thinking, and facilitation of associative/implicit processes, though findings vary across individuals and contexts. This literature motivates the proposal that AI systems can mimic psychedelic mechanisms to support creativity.

Perspective and conceptual synthesis. The author conducts a narrative integration of (a) psychedelic neuroscience/psychology (DMN modulation, set and setting, creativity effects) and (b) creativity science (latent inhibition, divergent/convergent thinking, implicit learning), then maps these mechanisms to AI design patterns for creativity augmentation. The paper illustrates potential implementations via existing tools (e.g., Jukebox for novel musical inputs, DeepDream for amplified pattern detection, GPT-4/Midjourney for associative recombination and reframing) and proposes future design features such as dynamic "conceptual mutation" layers to inject controlled novelty, affect-aware prioritization, and adaptive mentoring that personalizes outputs over time. It further outlines personality-sensitive tailoring (e.g., modulating semantic distance/ambiguity based on openness or conscientiousness) and suggests empirical paradigms (manipulating AI novelty range by user openness, measuring creative outputs and comfort) and neuroimaging studies (examining DMN-related effects during AI interaction). No primary empirical data are collected.

- Mapping psychedelic mechanisms to AI: (1) Reduce latent inhibition by bypassing user filters and introducing unfiltered, cross-domain inputs; (2) Enhance divergent thinking via cross-domain recombination and controlled disruption (e.g., conceptual mutation layers) that surface low-probability associations; (3) Strengthen implicit learning through adaptive, long-term interactions that scaffold pattern internalization. - Three design principles for creativity-enhancing AI: (i) Introduce controlled noise/novelty to disrupt mental filters; (ii) Generate cross-domain outputs with affective salience (e.g., surprise, aesthetic dissonance); (iii) Embed adaptive feedback loops to foster cumulative, intuitive insight. - Personality- and context-adaptive AI: Systems should tailor novelty intensity, semantic distance, associative breadth, emotional tone, and ambiguity to user traits (e.g., openness, conscientiousness) and state (set and setting), with real-time recalibration from explicit feedback or affective cues. - Risks and mitigations: Potential AI dependency and overreliance; homogenization and bias from training data (and recursive model collapse when trained on AI outputs); need for diverse datasets and inclusive oversight; preserve human-led convergent evaluation and decision-making. - Practical exemplars: AI can reframe problems (e.g., GPT-4), surface ignored patterns (DeepDream), and deliver novel inputs (Jukebox, Midjourney) to catalyze idea generation without claiming human-like creativity.

By aligning AI system design with mechanisms implicated in psychedelic-induced creativity—loosening cognitive constraints, broadening associative access, and supporting unconscious pattern learning—the paper proposes a pathway for AI to act as a cognitive catalyst. This addresses the central question of how AI can work with human psychology: design for controlled disruption (akin to DMN modulation), context dependence (set and setting), and individual differences (personality), while maintaining human control over convergent selection and ethics. The significance lies in enabling more flexible, exploratory, and insight-rich cognition across domains (education, design, therapy) and potentially democratizing access to creative augmentation. However, the approach requires safeguards against dependency, bias amplification, and output homogenization, and calls for empirical validation and neurocognitive measurement to ensure intended effects and equitable access.

The paper advances a conceptual framework for creativity-enhancing AI inspired by psychedelic neuroscience and creativity research. It identifies three mechanism-to-design mappings (latent inhibition reduction, divergent thinking stimulation, implicit learning support) and distills them into design principles (controlled novelty, affect-salient cross-domain generation, adaptive feedback). It argues for personality- and context-sensitive AI to broaden accessibility and efficacy. Future research directions include controlled user studies manipulating AI novelty and idea-space breadth across personality profiles (e.g., openness), neuroimaging to probe DMN-related and other neural effects during AI interaction, evaluation of ethical safeguards for implicit influence, dataset diversification to reduce bias and homogenization, and exploration of brain–computer interface pathways that respect user agency.

- Conceptual perspective without primary empirical data; claims require experimental validation. - Effects of psychedelics on creativity are heterogeneous and context-dependent; analogies to AI may not generalize across users or tasks. - Potential for AI dependency and reduction in critical thinking if overused; requires literacy and workflow designs that keep humans in control of convergent evaluation. - Risk of homogenization and bias due to training data composition and recursive training on AI outputs; inclusivity and data diversity are necessary but challenging. - Personality-adaptive systems may inadvertently widen disparities if adoption or comfort varies by trait; careful scaffolding and equity-focused design are needed. - Ethical concerns around implicit influence and affective/physiological monitoring require transparency, consent, and strong governance.