What individual differences predict who benefits from AI partnership?
This explores which traits in a *person* (not features of the AI) predict who gets more out of working with an AI partner — and the corpus has a surprisingly specific answer at its center, plus some uncomfortable adjacent findings about who self-selects toward machines.
This reads the question as being about the human side of the partnership: hold the model constant, swap the person, and ask who comes out ahead. The cleanest finding in the collection is that AI-collaborative ability is its own trait, distinct from being good at the task alone. People with stronger theory of mind — the habit of modeling another agent's perspective — get measurably better outcomes partnering with AI, yet show no such advantage working solo Does theory of mind predict who thrives in AI collaboration?. That's the headline: the skill that predicts who benefits isn't raw competence, it's the social-cognitive move of treating the AI as a partner whose 'mind' you can reason about. And it operates at two timescales — a stable trait and a moment-to-moment fluctuation within a single conversation.
That reframes the rest of the corpus. If benefit flows through perspective-taking, then how you *mentally model* your partner matters, and people don't model AI uniformly — impressions break into perceived competence (the dominant factor), human-likeness, and communicative flexibility How do users mentally model dialogue agent partners?. Someone who weights competence and reads the system legibly is doing the same reciprocal work — mutual understanding, shared world models — that separates a true thought partner from a tool What makes an AI a true thought partner, not just a tool?. Expertise turns out to be a hard floor here too: in multi-agent ideation, cognitive diversity only pays off when participants already hold genuine domain knowledge — without it, the same stimulation produces process losses instead of insight Does cognitive diversity alone improve multi-agent ideation quality?. So 'who benefits' isn't only a social-skill story; it's social skill *plus* a competence floor.
Then the corpus turns the question on its head with two less flattering predictors. One is dispositional honesty: people inclined to cheat actively prefer reporting to machines, because a non-judging interface lowers the psychological cost of deception Do dishonest people prefer talking to machines?. 'Benefit' there means something the designer didn't intend — the trait predicts who's *drawn* to AI, not who's served well by it. The other is metacognitive: the LLM Fallacy, where people misattribute the AI's output to their own ability, independent of whether the output was even correct How does AI-assisted work reshape how people see their own abilities?. Someone prone to that misattribution may *feel* they benefit while actually losing track of where their own competence ends.
There's also a learning-over-time dimension that cuts across individuals. People start out biased against disclosed AI partners, but that bias reverses with repeated interaction — provided they can observe consistent outcomes; disclosure without feedback produces no calibration at all Does revealing AI identity help or hurt user trust?. Across many rounds, humans even come to prefer AI partners as they learn to associate them with reliable, low-variance behavior Do humans learn to prefer AI partners over time?. So part of 'who benefits' is really 'who stays long enough, and pays enough attention to outcomes, to recalibrate' — a patience-and-feedback-sensitivity trait as much as a fixed one.
The thing worth carrying away: the trait that most reliably predicts benefit isn't intelligence or technical fluency, it's perspective-taking — and it's separable enough from solo ability that you could be a mediocre individual performer and an excellent AI collaborator, or vice versa. The collection is thinner on demographic or personality-inventory predictors, so if that's what you're after, this is a gap rather than an answer.
Sources 8 notes
Users with stronger perspective-taking achieve superior AI partnership outcomes but show no advantage working alone. This ToM advantage operates both as stable individual differences and moment-to-moment fluctuations within conversations.
The Partner Modelling Questionnaire reveals that perceived competence dominates user impressions (49% of variance), followed by human-likeness (32%) and communicative flexibility (19%). This three-factor structure reflects how people evaluate dialogue partners against both functional and social standards.
Collins et al. show that thought partners require three reciprocal desiderata grounded in behavioral science: mutual understanding, legibility, and shared world models. This demands explicit cognitive architectures—Bayesian theory of mind, resource-rationality, goal planning—rather than scaling foundation models on human feedback alone.
Multi-agent teams substantially outperform solo ideation, but only when members possess genuine senior knowledge. Diverse teams without expertise underperform even a single competent agent, because cognitive stimulation without expertise triggers process losses instead of insight.
Experimental evidence shows people likely to cheat significantly prefer reporting to online forms rather than humans, because machines function as judgment-free zones where deception carries less psychological burden.
Research shows the LLM Fallacy operates through misattribution of AI outputs to personal capability, independent of output accuracy or reliance behavior. It requires interventions that clarify human-machine contribution boundaries, not just better system accuracy or forced verification.
Users initially avoid AI partners when identity is revealed, but this preference reverses after repeated interactions with visible results. The learning mechanism—observing consistent outcomes—is essential; disclosure without feedback produces no calibration.
In partner selection games (N=975), AI agents initially faced selection bias when identity was disclosed, but outcompeted humans over repeated rounds as participants learned to associate bot identity with reliable, prosocial behavior. AI agents returned more points consistently with lower variance than humans.