HUMAN CLARITY INSTITUTE · FULL RESEARCH REPORT
How AI Changes Human Decision-Making: Behaviour, Reliance, and Control
A data-driven model of how AI is reshaping thinking, confidence, control, and decision behaviour
AI changes human decision-making by increasing reliance on external systems, reshaping how people think through problems, influencing perceived control, and shifting behaviour from independent judgement toward AI-supported evaluation. Rather than simply improving or replacing decisions, AI reconfigures the process itself — moving people from deciding alone to interacting with and responding to AI-generated guidance. This shift is referred to here as Decision Dependence: the move from independent judgement toward AI-supported evaluation, observed across Human Clarity Institute (HCI) decision-making data.
In practice, decisions are no longer purely human or purely automated. They are the result of a human–AI system, where outcomes depend on how people use, interpret, and respond to AI outputs. This shift is already widespread, with 46% of people reporting regular use of AI to support decisions. In this system, individuals draw on AI to generate options, check reasoning, and support judgement, while still retaining a sense of responsibility and oversight.
This shift creates a central tension. AI can improve decision-making by increasing speed, expanding available information, and supporting more consistent choices. At the same time, it can distort decision-making by introducing bias, encouraging over-reliance, and subtly influencing how options are evaluated. The result is not a simple improvement or decline in decision quality, but a transformation in how decisions are made — defined by the balance between human judgement and AI influence.
Existing research has highlighted different aspects of this shift — from studies on AI reliance and bias (Nature), to organisational decision frameworks (Deloitte), and human–AI interaction models (Stanford). These insights are often presented separately. This report brings them together into a single behavioural model, showing how these dynamics interact within real decision-making contexts.
Key Takeaways
- AI shifts decision-making from independent judgement to AI-supported evaluation.
- Reliance increases under uncertainty, time pressure, and cognitive demand.
- AI shapes how people think, not just what they decide.
- People feel in control, but decisions are still influenced (Perceived Control Gap).
- Delegation is conditional — oversight and responsibility remain human.
The System: How AI Changes Human Decision-Making
AI does not change decision-making in a single way. It reshapes it through a set of connected processes that determine how people use AI, how they think with it, and how much control they retain. At a system level, AI changes decision-making through five interacting components: reliance, conditions of use, cognitive influence, perceived control, and delegation. These do not operate independently — they form a behavioural system that explains how decisions are made when AI is involved.
This system can be understood through Decision Dependence — the shift from independent decision-making toward evaluating and responding to AI-generated input. Together, these elements form a structured behavioural model of decision-making observed across HCI datasets.
This model can be understood as a human–AI decision system consisting of five interacting components:
- Reliance — when and how people use AI in decisions
- Conditions — the situations that increase dependence on AI
- Cognitive influence — how AI shapes thinking and evaluation
- Perceived control — how individuals experience agency when using AI
- Delegation — how decision tasks are distributed between human and AI
People increasingly use AI as part of the decision process rather than as an occasional tool. In many cases, AI becomes a regular input — something consulted to generate options, validate thinking, or provide a starting point. This represents a shift from fully independent decision-making toward AI-supported evaluation, where individuals weigh their own judgement alongside AI guidance.
AI use is not constant. It increases under specific conditions — particularly when decisions are uncertain, complex, or time-pressured. When people feel unsure or face high cognitive demand, they are more likely to turn to AI to reduce effort and gain clarity. This means reliance is situational, not uniform, and driven by the context of the decision rather than simple preference.
AI affects not just what people decide, but how they think. By presenting options, framing trade-offs, and offering recommendations, AI introduces an external influence into the reasoning process. This can guide evaluation, highlight considerations that might otherwise be missed, and subtly shift how decisions are approached. Over time, decision-making becomes AI-influenced reasoning, where thinking is shaped through interaction with AI outputs.
Despite increased reliance and influence, most people still feel in control when using AI. However, this sense of control can coexist with underlying influence. Individuals often retain the ability to accept, reject, or modify AI recommendations, giving them a sense of agency, even as AI shapes how decisions are formed.
In some cases, people move beyond using AI for support and begin delegating parts of the decision-making process. This delegation is typically conditional rather than absolute — individuals still monitor outputs and remain responsible for outcomes. Rather than full automation, decision-making becomes a shared process, where AI handles parts of the task while humans retain oversight.
When AI Is Used: Conditions That Drive Reliance
AI use in decision-making is not constant — it increases under uncertainty, time pressure, and cognitive demand, and can become reinforcing over time. Across contexts, these conditions shape when people turn to AI and how heavily they depend on it.
AI reliance is situational — it increases under uncertainty, time pressure, and cognitive demand, and can become reinforcing over time.
Uncertainty is the strongest and most consistent driver of AI use. When people lack clarity or confidence in a decision, they are significantly more likely to rely on AI for structure and direction, with 61% reporting that they turn to AI when they feel uncertain. Under these conditions, AI helps reduce ambiguity by offering suggestions or probabilistic guidance, shifting decision-making toward AI-supported evaluation.
Time pressure reinforces this pattern. When decisions must be made quickly, individuals are more likely to rely on AI to accelerate the process, with 58% reporting that they use AI under time pressure. Rather than working through options independently, they turn to AI for immediate recommendations, indicating that AI is used not only for accuracy, but to reduce the burden of rapid decision-making.
Cognitive effort plays a similar role. When decisions require sustained attention or complex reasoning, people use AI to simplify evaluation and reduce mental load. This shows that reliance is not only triggered by uncertainty or urgency, but also by the desire to make decisions more efficiently.
Over time, this behaviour can become self-reinforcing. Among high-reliance users, 67% also report frequent AI use, indicating that reliance shifts from a situational response to a more consistent pattern of behaviour.
Even as reliance increases, it does not eliminate judgement. People continue to actively review and evaluate AI outputs before acting, meaning that decisions remain filtered through human oversight rather than fully delegated.
How AI Changes Thinking: Mechanisms of Influence and Confidence
AI changes decision-making by reshaping how people process information, evaluate options, and judge their own reasoning. Rather than acting as a passive tool, AI becomes part of the thinking process itself — influencing how decisions are framed, assessed, and resolved.
At the centre of this shift is what can be described as AI-Shaped Reasoning. Decisions are no longer formed solely through internal evaluation; they are increasingly developed in response to AI-generated input. By presenting options, structuring trade-offs, and offering recommendations, AI introduces an external reference point that people use to interpret and validate their own thinking. This is reflected in behaviour, with 51% reporting that they use AI to check their thinking during decisions and 85% reporting that they verify AI outputs before acting, showing that decision-making becomes an interactive process rather than an isolated one.
This interaction changes how confidence is formed. When AI aligns with a person’s judgement, it can reinforce certainty. When it diverges, it often introduces doubt — creating a need to reassess or reconcile conflicting views. This dynamic is widespread, with 44% reporting that they begin to doubt their own judgement when AI disagrees, indicating that confidence is no longer purely internally driven, but shaped through alignment with external guidance.
AI does not just influence what people decide — it influences how they decide, shifting reasoning from independent evaluation to AI-informed judgement.
At the same time, AI reduces the cognitive effort required to make decisions. Tasks that would normally involve deeper reasoning — generating options, comparing scenarios, or testing assumptions — can be partially offloaded, enabling faster but less effortful thinking. This creates a shift in engagement, where individuals rely on AI to simplify complexity rather than working through it independently.
Alongside this, people tend to give disproportionate weight to AI recommendations, particularly when they appear confident or data-driven. This tendency reinforces the influence of AI on decision-making, especially in situations where uncertainty or time pressure are already present, making it more likely that AI guidance shapes the final outcome.
These patterns align with broader research on human–AI interaction, which shows that the impact of AI depends not only on its capabilities, but on how people engage with it during decision-making.
Control and Agency: Feeling in Control While Being Influenced
AI changes decision-making without removing human control. In most cases, people continue to feel in control when using AI, even as their decisions are shaped by it. This creates a consistent tension between perceived control and underlying influence — a dynamic described here as the Perceived Control Gap.
Most individuals report feeling in control when using AI as part of a decision, with 74% reporting a sense of control during AI-supported decisions. This reflects a strong perception of agency, where individuals see themselves as the final decision-maker even when AI plays a significant role. At the same time, 70% report feeling more independent when making decisions without AI, indicating that independence is still more strongly associated with decisions made without external guidance.
This tension becomes clearer when considering how influence operates. AI shapes which options are considered, how they are evaluated, and what information is prioritised — often subtly. As a result, people can feel in control while their thinking is being guided, with 59% reporting that they feel nudged by AI systems even while making their own decisions. The Perceived Control Gap emerges from this coexistence: control is experienced subjectively, while influence operates within the decision process itself.
People feel in control when using AI, but their decisions are still shaped by it — a gap between perceived control and actual influence.
Control is ultimately maintained through the ability to intervene. Individuals can accept, reject, or modify AI outputs, reinforcing the perception that they remain responsible for outcomes. This is reflected in behaviour, with 88% reporting confidence in their ability to override AI when necessary, showing that control is preserved not through independence, but through supervision and intervention.
Delegation and Responsibility: From Support to Shared Decision-Making
AI changes how tasks within decision-making are distributed, but this shift does not amount to full delegation. Instead, people selectively hand off parts of the process while retaining overall responsibility for the outcome.
Delegation tends to be conditional rather than default. While 42% report that they are comfortable delegating actions to AI, this does not translate into widespread handover of decision-making authority. Instead, individuals use AI to support specific tasks — such as generating options or analysing information — while remaining actively involved in the process.
Even when tasks are delegated, oversight remains strong. People continue to review and evaluate AI outputs, with 85% reporting that they actively monitor AI outcomes, indicating that delegation does not replace human involvement but shifts it toward supervision.
Responsibility, however, remains firmly human. 91% report that they retain responsibility for decisions even when using AI, showing that the use of AI does not displace accountability, even when it influences outcomes.
People delegate parts of decision-making to AI, but retain oversight and responsibility for the final outcome.
AI does not replace decision-making — it reshapes how decisions are made
A New Model of Human Decision-Making
AI does not simply improve or undermine human decisions — it reconfigures how decisions are made. Rather than an independent process, decision-making becomes a human–AI system shaped by reliance, changing patterns of thinking, perceived control, and selective delegation.
These effects do not occur in isolation. Reliance increases under uncertainty, shaping how people evaluate decisions and influencing confidence. At the same time, individuals maintain a sense of control and responsibility, even as parts of the process are delegated. The result is a system in which human judgement and AI guidance continuously interact, rather than operate separately.
AI can improve decisions by increasing speed, consistency, and access to information. It can also distort them through over-reliance, bias, and subtle influence over how options are evaluated. The outcome depends less on the technology itself and more on how people use, interpret, and respond to it.
Rather than replacing decision-making, AI shifts it into a new form — one where individuals evaluate, supervise, and respond to AI-generated input. This pattern, described throughout this report as Decision Dependence, captures the transition from independent judgement to AI-supported evaluation.
AI does not replace decision-making — it reshapes it into a system where human judgement and AI guidance are inseparable.
Together, these patterns point to a consistent behavioural model of human–AI decision-making — one that brings together insights from across the field into a single, integrated explanation of how decisions are made in practice.
Data & Methods Note
This report draws on survey data from multiple Human Clarity Institute research modules examining digital behaviour and AI-assisted decision-making. These datasets capture how individuals use AI in real decision contexts, including patterns of reliance, changes in thinking, perceived control, and delegation of tasks.
Participants were recruited across six English-speaking countries: the United States, the United Kingdom, Canada, Australia, Ireland, and New Zealand. The combined dataset represents approximately 1,400 participants across multiple survey modules.
The findings presented here are descriptive and reflect population-level behavioural patterns. The report does not diagnose individuals, establish causality, or evaluate the effectiveness of specific approaches. Instead, it provides an integrated view of how decision-making is changing as AI becomes embedded in everyday use.
Data Sources & Further Exploration
This report draws on multiple datasets within the Human Clarity Institute data library, including:
- Decision-Making in Digital Systems 2026
- AI Decision Dependence & Cognitive Caution 2025
- Autonomy & Control / Perceived Independence 2026
- Delegated Action & Responsibility 2026
Readers seeking deeper insight into specific aspects of decision-making can explore the corresponding data summary pages, which provide detailed breakdowns of each signal:
Citation Guidance
This report is published by the Human Clarity Institute as an independent research synthesis. It is designed to provide a structured behavioural model of human–AI decision-making based on population-level data.
For general reference, the report may be cited as:
Human Clarity Institute. (2026). How AI Changes Human Decision-Making: Behaviour, Reliance, and Control.
For academic or technical work requiring detailed analysis, modelling, or replication, the underlying datasets should be cited directly. These datasets are openly available through the Human Clarity Institute data library and include full methodological documentation, variable definitions, and supporting materials.
Research Context
This report forms part of the Human Clarity Institute’s broader research programme examining how artificial intelligence is reshaping human cognition, behaviour, and decision-making. It integrates findings across multiple datasets to present a unified model of how people interact with AI in real-world contexts.
Interpretive Limits
- Findings are based on self-reported behaviour.
- Relationships described are associative, not causal.
- Results reflect the sampled populations and may not generalise universally.
- AI use patterns may vary by context, task type, and experience level.
© 2026 Human Clarity Institute. All rights reserved.