This is the one page in the WILD ecosystem where the research stands onstage. Everywhere else, wisdom leads and the studies stay backstage, in service of a reader-path; here, for the reader who wants to see the studies themselves, they are the whole point.
Every claim below is graded by how strong the evidence actually is - peer-reviewed journal, preprint, or working paper - and scoped to exactly what it shows, because overclaiming a fragile finding is itself a failure of the wisdom this page is describing. Organized in the framework's own order: Wisdom, Intention, Leadership, Discovery.
Wisdom
The science of judgment under uncertainty - the load-bearing cluster: the empirical case that wisdom is distinct from intelligence, and trainable.
The construct
The field converged (Grossmann et al. 2020, the “Common Wisdom Model”) on wisdom as perspectival metacognition - intellectual humility, context-sensitivity, integration of perspectives, awareness of uncertainty and change - wedded to a moral orientation toward a balance of interests. It is operationalized and scoreable (the Berlin Wisdom Paradigm, Baltes & Staudinger 2000; the Situated Wise Reasoning Scale, Brienza et al. 2018).
Distinct from IQ, and the better predictor of what matters
Wisdom is psychometrically distinct from intelligence (Staudinger, Lopez & Baltes 1997); wise reasoning rises into old age even as fluid IQ falls (Grossmann et al. 2010, PNAS); head-to-head, wise reasoning predicts life satisfaction, relationship quality, and lower negative affect while intelligence does not, sometimes running the wrong way (Grossmann et al. 2013, JEP: General).
Trainable as a stance
Wise reasoning varies more within a person across situations than between people. Self-distancing raises it (Kross & Grossmann 2012). Solomon's paradox - we reason more wisely about others' problems than our own - and self-distancing erases the gap (Grossmann & Kross 2014). A preregistered month-long diary of distanced self-reflection increased wise reasoning, N=555 (Grossmann et al. 2021).
The scoped claim
“Wisdom matters more than intelligence” is well-founded under uncertainty and for flourishing; for well-defined, speeded problems, IQ still wins. That scope is the AI-age argument, and it stays a scoped claim rather than a universal slogan (the universal version is Sternberg's theoretical argument, not data).
How strong is this
Distinct-from-IQ and predicts-flourishing: robust, multi-paradigm. Trainability: strong but young (~1-month follow-ups). Universal wisdom-over-IQ: theoretical (Sternberg), flagged as such. Honest caveat: wisdom science has real measurement debates - low convergent validity across instruments, and a self-report paradox, since the genuinely wise tend to rate themselves lower. Which is why wisdom is measured in situations, never by self-rating.
Intention
The science of purpose and self-authored direction.
The construct
Purpose is a self-organizing aim that generates and prioritizes goals, distinct from any single goal (McKnight & Kashdan 2009); self-concordant goals, chosen for identity-congruent reasons, draw more effort and yield more well-being than controlled ones (Sheldon & Elliot 1999).
Compounding and trainable
Self-concordant striving feeds an upward spiral (Sheldon & Houser-Marko 2001); purpose responds to structured exercises (Bronk) and to values / best-possible-self writing (King 2001; Cohen & Sherman 2014). Hard outcome data, rare for a “soft” construct: higher purpose predicts roughly 15% lower 14-year mortality per standard deviation (Hill & Turiano 2014), confirmed across 136,265 people (Cohen, Bavishi & Rozanski 2016).
The threat
When AI supplies the heading, that is the overjustification effect - an externally furnished motive crowding out the self-authored one (Deci 1971; Deci, Koestner & Ryan 1999, a 128-experiment meta-analysis) - compounded by automation deference (Logg et al. 2019).
How strong is this
Self-Determination Theory, the overjustification effect, and self-concordance: robust, decades of replication. Purpose-mortality: strong but observational and modest. Interventions: promising, younger. Honest note: the goal-setting debate is live - “Goals Gone Wild” (Ordóñez et al. 2009) is the failure mode of over-optimizing a fixed destination, not proof that goals themselves are bad (Locke & Latham rebut it).
Leadership
The science of productive difficulty and optimal challenge.
The mechanism
Durable learning lives in storage strength, bought only with effortful retrieval (Bjork & Bjork; “desirable difficulties,” Bjork 1994). The workhorse effects are among the most replicated in cognitive psychology: the testing / retrieval effect (Roediger & Karpicke 2006), the generation effect (Slamecka & Graf 1978), the spacing effect (Cepeda et al. 2006). Even failed retrieval before the answer helps (Kornell, Hays & Bjork 2009). Optimal challenge is a moving channel (Csikszentmihalyi's flow; Vygotsky's zone of proximal development).
The threat
Outsource the struggle, lose the storage. The literal case for the navigation metaphor: heavier lifetime GPS use predicts worse hippocampal spatial memory and a steeper decline over three years (Dahmani & Bohbot 2020). The offloading principle is established (Sparrow et al. 2011; Risko & Gilbert 2016).
How strong is this
Testing, generation, and spacing effects: robust, decision-grade. Desirable difficulties: strong organizing theory. Flow and the zone of proximal development: foundational but loosely operationalized. Honest limit: deliberate practice is often overstated - Macnamara et al. (2014, 2016) put it at roughly 26% of variance in games, 4% in education, under 1% in professions. The honest version: effortful practice is necessary, not sufficient, which strengthens the case for Leadership, since practice does not run on autopilot and steering it is the scarce skill. LLM-specific “cognitive debt” work (Kosmyna 2025) is early-signal preprint only.
Discovery
The science of curiosity and exploration.
The mechanism
Curiosity is an information gap, a drive-like, aversive state that fires when you perceive a hole in what you know (Loewenstein 1994). The gap recruits dopaminergic reward circuitry and strengthens hippocampal encoding, so curiosity makes learning stick (Gruber, Gelman & Ranganath 2014; the PACE framework, Gruber & Ranganath 2019). It peaks at the edge of competence, the inverted-U (Kang et al. 2009).
Trainable
A meta-analysis of 41 randomized controlled trials (4,496 participants) puts curiosity interventions at Hedges' g = 0.57 (Schutte & Malouff 2023). There is a trait substrate (CEI-II; Kashdan et al. 2009) and a movable state.
The threat
Answer-on-demand technology closes the gap before it opens: it reduces willingness to attempt questions (Ferguson et al. 2015), inflates a false sense of one's own knowledge (Fisher et al. 2015), and offloads memory outward (Sparrow et al. 2011).
How strong is this
The mechanism - gap, dopamine, memory - is robust. Trainability is strong but heterogeneous. Honest flag: the “AI kills curiosity” claim is search-engine-era and measures memory and overconfidence, not a longitudinal fall in trait curiosity. It is a grounded extrapolation, not a demonstrated result - the open frontier, named as such.
Design decides
The empirical floor under “amplifier, not crutch.”
The same technology builds or hollows depending on design. A raw chatbot that hands over answers boosts practice but leaves users worse on an unassisted test than peers who never used it; the same model behind a prompt that withholds the answer and coaches instead erases nearly all of the harm (Bastani et al. 2025, PNAS). A purpose-built, guardrailed tutor more than doubled the learning gains of a strong classroom (Kestin et al. 2025, Scientific Reports). Self-Determination Theory names why it holds: motivation grows where autonomy, competence, and relatedness are met rather than outsourced (Deci & Ryan).