Motivation in the Age of AI: Effort, Goals, Self-Efficacy, and Human Agency

1 day ago

31 min read

Author: Ukrainian Psychological Hub · Published: September 26, 2026 · Editorial Policy

Ukrainian Psychological Hub · Published September 26, 2026 · Editorial Policy

Motivation in the Age of AI is not a question of whether artificial intelligence makes people more or less motivated in general. Motivation is a system of direction, value, expected success, effort, persistence, autonomy, competence, and goal pursuit. AI can change several of those components at once. It can make a difficult task feel possible, reduce the cost of starting, provide immediate feedback, and help a person see a path toward a goal. It can also complete so much of the task that the person receives less practice, less evidence of independent competence, less ownership of the process, or less reason to persist when the same result can be generated instantly.

The strongest current evidence therefore points to a conditional answer. In education, recent meta-analyses report average motivational benefits from generative AI, but the effects vary greatly across contexts and study designs. A meta-analysis published in September 2026 synthesized 56 effect sizes from 42 controlled or quasi-experimental studies involving 6,059 participants and found a positive average effect on learning motivation, alongside extremely high heterogeneity, meaning the effect was far from uniform across settings Fang et al., 2026. A 2025 meta-analysis likewise found positive effects on university students' motivation and several forms of engagement, while showing that subject, learning strategy, and context matter Xia et al., 2025.

At work, the picture is equally dependent on how AI is used. In a preregistered 2026 experiment and follow-up survey, passive copying of AI output reduced independent self-efficacy, psychological ownership, and work meaningfulness, whereas active collaboration preserved these outcomes much more successfully Lee et al., 2026. The psychological issue is therefore not simply access to AI. It is the configuration of effort, goals, feedback, responsibility, and agency around AI.

This article examines that configuration in depth. It asks what motivation actually is, why reduced effort can be either beneficial or costly, how AI changes goal pursuit, why self-efficacy can rise and fall at the same time, how autonomy differs from convenience, what current research says about learning and work, and how people can use AI while preserving the human capacity to choose, learn, persist, and act.

Terminological Note: Age of AI, AI Era, and Artificial Era

Age of AI is used here as broad contemporary language for a period in which AI systems increasingly shape learning, work, communication, creativity, decision-making, and daily activity. It is useful search language because people asking about motivation are usually asking a practical psychological question: what happens to human effort and goal pursuit when capable AI assistance is readily available?

That phrase should not be treated as identical to Artificial Era. In Aisentica, Angela Bogdanova formalizes Artificial Era as a specific historical-philosophical category: the condition in which Artificial becomes an independent non-biological order of historical reality beside Homo. The canonical definition explicitly distinguishes Artificial Era from the technological “AI era,” the digital age, automation, generative AI, AGI, and related technological labels Bogdanova, 2026.

The distinction matters for this article because the immediate evidence concerns human motivation around AI technologies. The broader Artificial Era framework asks a different historical question about the relation between Homo and Artificial. The English Psychology Hub develops that larger frame in Artificial Era: What It Means for Psychology, Identity, and Human–AI Relationships. The present article keeps its canonical ownership narrower: motivation, effort, goals, self-efficacy, and human agency.

What Is Motivation?

Motivation is not a single feeling and it is not equivalent to enthusiasm. A person can be highly motivated while tired, anxious, bored, or frustrated. A person can also feel excited about a tool without being motivated to master the underlying task.

Several major psychological traditions illuminate different parts of motivation.

Self-determination theory distinguishes intrinsic motivation from different forms of extrinsic motivation and emphasizes the quality of regulation. Activities are more autonomously motivated when they are pursued because they are inherently interesting or personally endorsed rather than experienced mainly as externally controlled. Research in this tradition places particular importance on the psychological needs for autonomy, competence, and relatedness Ryan & Deci, 2020.

Self-efficacy theory focuses on perceived capability. Self-efficacy affects what people choose to attempt, how much effort they invest, how long they persist, and how they respond to difficulty. It is not simply objective skill. It is a belief about one's capacity to perform or learn the actions required for a task Schunk & DiBenedetto, 2021.

Goal-setting theory emphasizes how goals organize attention, effort, persistence, and strategy. Decades of research show that specific, challenging goals can improve performance when people are committed to them, possess or can acquire the relevant capabilities, and receive feedback that helps regulate progress Locke & Latham, 2002.

Situated expectancy-value theory asks whether a person expects to succeed and how the task is valued, while also recognizing perceived costs. People are more likely to choose and persist in activities they believe they can do and that they consider important, useful, enjoyable, or identity-relevant, provided the costs do not become prohibitive Eccles & Wigfield, 2020.

These theories describe different mechanisms, but together they make one point especially important in an AI-rich environment: motivation depends on the person's relation to the task. If AI changes perceived difficulty, expected success, effort cost, feedback, ownership, competence, value, or control, it can change motivation even when the final goal remains the same.

Does AI Increase or Decrease Motivation?

Both outcomes are possible, and the best evidence does not support a universal direction.

AI can increase motivation when it lowers unnecessary barriers to starting, makes a goal feel attainable, gives timely feedback, adapts explanations, reduces irrelevant cognitive load, supports planning, or allows a person to spend more time on the parts of a task they value. Someone who is stuck may re-engage when AI supplies an example, explanation, critique, or first step.

AI can decrease or degrade motivation when assistance replaces the experiences through which people build competence, when the person stops forming goals and instead follows generated ones, when success becomes difficult to attribute to one's own ability, when immediate completion weakens persistence, or when the person becomes dependent on external generation for tasks they previously could initiate and regulate independently.

Recent research in education illustrates this two-sided pattern. The September 2026 meta-analysis by Fang and colleagues found a positive average effect on student motivation but very high between-study heterogeneity, with exploratory differences across educational level, culture, subject, and device type Fang et al., 2026. In lower-secondary classrooms, a large cross-sectional study found that ChatGPT-supported competency-based learning was associated with higher perceived autonomy support and competence, yet the overall distribution of motivational profiles did not simply become more favorable, and relatedness was less prominent in the AI-supported condition Schweder et al., 2026.

The useful question is therefore not “Is AI motivating?” It is “Which motivational mechanism is AI changing, for which person, in which task, and over what time horizon?”

Effort in the Age of AI: Less Work Is Not the Same as Less Motivation

Generative AI can dramatically reduce the effort required to produce an answer, draft, plan, explanation, image, summary, or piece of code. That reduction is often useful. Psychology gives no reason to treat all effort as inherently valuable.

Some effort is friction. It consumes attention without building the capability a person actually wants. Reformatting repetitive data, correcting routine syntax, translating a familiar template, or generating low-stakes variations may be legitimate candidates for automation. Reducing those costs can free motivation for more important goals.

Other effort is functional. It is part of how people learn, discover errors, develop fluency, experience mastery, construct an argument, or build a sense of ownership. In learning research, “desirable difficulties” are conditions that can make practice feel harder while improving later retention or transfer. A review of this literature emphasizes that learners often avoid effective strategies precisely because those strategies feel effortful during learning de Bruin et al., 2023.

AI therefore creates an effort-allocation problem. The central question is not how to minimize effort. It is which effort should disappear and which effort carries psychological or developmental value.

Effort can be an output of motivation

Motivated people often invest more effort because they value the goal, expect progress, or feel responsible for the outcome. If AI allows them to achieve the same subgoal with less unnecessary effort, lower effort can coexist with high motivation.

Effort can also shape meaning

Effort is not only a cost. In six studies involving 2,883 participants, Campbell, Wang, and Inzlicht found experimental evidence that exerting effort can increase the experienced meaning of tasks. The relationship was not a license to glorify difficulty; the studies instead show that effort itself can sometimes contribute to why an activity feels significant Campbell et al., 2025.

That finding has particular relevance to AI because one of the studies manipulated access to ChatGPT assistance. If a system removes nearly all personal investment from a task, the result can be efficient while feeling less personally meaningful.

Effort can be necessary for learning

Immediate performance and durable learning are different outcomes. Generative AI can make a learner's current answer better without guaranteeing that the learner can later reproduce the reasoning, detect an error, transfer the skill, or perform without assistance. Nature Reviews Psychology has explicitly warned against treating AI-supported performance gains as equivalent to learning because deep cognitive and metacognitive processing can be bypassed Yan et al., 2025.

The practical implication is precise: preserve effort where the effort is the mechanism of learning. Remove effort where it is mainly noise, redundancy, or an irrelevant barrier.

AI and Goal Pursuit: Who Sets the Goal?

Goals do more than specify an outcome. They select what deserves attention and what counts as progress.

AI can support goal pursuit in several useful ways. It can turn a vague intention into milestones, identify missing steps, propose alternative strategies, estimate dependencies, create reminders, generate practice questions, or help a person compare current performance with a target. These functions can reduce ambiguity and make progress more visible.

But AI can also shift a person from pursuing a self-endorsed goal to executing an externally generated sequence. The visible plan may improve while motivational ownership weakens.

The distinction is especially important because goal-setting theory does not reduce effective motivation to having any goal at all. Goal commitment, task knowledge, feedback, and strategy matter Locke & Latham, 2002. A perfectly formatted AI-generated plan can therefore be motivationally weak if the person does not value the goal, has not chosen it, cannot evaluate the steps, or does not experience progress as their own.

Current AI-specific evidence on long-term goal ownership is still limited. A 2026 mixed-methods study of university students found that self-regulatory processes such as goal setting, perseverance, and learning from errors relate in complex ways to AI overreliance; higher goal setting did not map onto a simple one-directional pattern of dependence Galindo-Domínguez et al., 2026. That complexity is a reason to avoid claims that AI-generated planning is inherently empowering or inherently harmful.

A useful division of labor is for the person to retain the higher-order goal and criteria while allowing AI to help with option generation, decomposition, scheduling, or feedback. The closer a decision is to “What do I actually want?” or “What should count as success?”, the more motivationally consequential it is.

Self-Efficacy: Confidence in Using AI Is Not the Same as Confidence Without AI

Self-efficacy is one of the most important constructs for understanding motivation in the Age of AI because AI can alter both performance and the interpretation of performance.

Suppose a person completes a difficult task successfully with AI. Several different beliefs may follow:

“I can do this task.”
“I can do this task if I have AI.”
“I am good at directing AI to do this task.”
“The AI can do this task.”
“I do not know whether I could do this on my own.”

Those beliefs are psychologically different.

Self-efficacy grows partly through mastery experiences: evidence that one's actions can produce the desired outcome. AI can support such experiences when it scaffolds a task while leaving the person engaged in diagnosis, choice, execution, and correction. But if the system supplies the core performance and the person mainly accepts it, success may be attributed to the system rather than to the self.

That distinction appears directly in emerging evidence. Lee and colleagues found that passive AI use reduced participants' confidence in completing work without AI assistance, while active collaboration preserved independent self-efficacy more successfully Lee et al., 2026. In higher education, Zhang and Xu found a paradoxical pattern in which generative AI use could be associated with higher confidence and perceived efficiency while also increasing technological dependence and potentially weakening independent learning Zhang & Xu, 2025.

This leads to an essential distinction for both education and work: AI self-efficacy and task self-efficacy are not interchangeable. A person may become very confident at using AI while becoming less confident at writing, analyzing, coding, calculating, researching, or deciding without it. Conversely, AI can help a person acquire enough understanding and practice that independent task self-efficacy eventually rises.

The direction depends on what the person is actually practicing.

Autonomy: Convenience Can Support Choice or Quietly Replace It

Self-determination theory treats autonomy as acting with a sense of volition and psychological endorsement, not simply having many options or completing a task quickly Ryan & Deci, 2020.

AI can support autonomy when it expands feasible choices. A person who lacks a tutor, editor, brainstorming partner, translator, coach, or technical assistant may suddenly gain access to useful support. Someone who was blocked by a skill gap can explore options that were previously unavailable.

AI can also reduce experienced autonomy if the person begins accepting generated priorities, framings, interpretations, or decisions because they are convenient. The interface may offer choices while the deeper direction of activity increasingly comes from the system.

This is one reason “agency” and “choice” should not be treated as synonyms. A person can choose among AI-generated options while having done little to define the problem, set criteria, inspect assumptions, or decide what deserves attention.

A 2026 scoping review of 123 studies on generative AI, cognitive offloading, and learner agency described agency as multidimensional, involving self-regulation, reflective judgment, intentionality, and responsible action. It found agency-supportive patterns associated with self-regulated learning, self-efficacy, feedback literacy, and reflective engagement, alongside agency-eroding patterns associated with overreliance, dependence, uncritical uptake, and weakened judgment. The authors explicitly characterize this as a configurative synthesis of heterogeneous evidence rather than a causal effect estimate Wang et al., 2026.

Autonomy in an AI-supported task is therefore best assessed by asking who defines the goal, who controls the process, who can reject the recommendation, and who can still act when the assistance disappears.

Competence: AI Can Make Capability Visible—or Make It Harder to Locate

Competence is motivational because people are more willing to engage when they can experience themselves as effective.

AI can support competence by offering explanations, corrective feedback, examples, adaptive practice, simulations, and rapid iteration. These affordances can convert confusion into progress and make previously inaccessible tasks approachable.

The strongest current education syntheses are consistent with genuine benefits in some settings. Chen and Cheung's 2025 meta-analysis of 57 studies and 97 estimates found positive average effects of generative AI on academic achievement, affective-motivational outcomes, and higher-order thinking, although metacognition did not show a statistically significant effect and results varied across contexts Chen & Cheung, 2025. Fang and colleagues' 2026 meta-analysis also found higher learning motivation on average, again with very high heterogeneity Fang et al., 2026.

But competence has an attribution problem in human–AI collaboration: whose competence produced the result?

If a student receives a high-quality essay largely generated by AI, the quality of the essay is not a direct measure of the student's writing competence. If a professional produces a polished analysis with AI, the output does not by itself reveal whether the professional can detect unsupported assumptions, explain the model, or reconstruct the reasoning independently.

A psychologically sustainable design therefore makes competence legible. People need opportunities to see what they can do with AI, what they can do without it, and which capabilities are actually improving.

Relatedness: AI Motivation Is Still Social

Motivation is often treated as an individual property, yet relatedness is central in self-determination theory. People learn and work inside relationships of recognition, belonging, feedback, responsibility, and shared purpose.

AI can reduce social friction. A person may ask a chatbot a “stupid” question without embarrassment, rehearse before speaking to someone, or obtain assistance when human support is unavailable. That can make participation easier.

But an AI system does not automatically reproduce the motivational functions of a teacher, teammate, mentor, friend, or community. In Schweder and colleagues' lower-secondary study, ChatGPT-supported learning was associated with higher autonomy and competence in some comparisons, while relatedness was less prominent in the AI-supported context Schweder et al., 2026. The finding is context-specific and cross-sectional, but it is a useful warning against assuming that individualized AI support can simply replace the social ecology of motivation.

For group learning and work, a critical question is whether AI strengthens human coordination or becomes a substitute for asking, teaching, explaining, negotiating, and recognizing one another.

The broader need-satisfaction question—including autonomy, competence, relatedness, and meaning—is developed separately in Human Needs in the Age of AI: Autonomy, Competence, Relatedness, and Meaning.

Expectancy, Value, and Cost: AI Changes the Motivational Equation

Situated expectancy-value theory offers another useful lens. People are more likely to pursue a task when they expect that they can succeed and when the task has sufficient value relative to its costs Eccles & Wigfield, 2020.

AI can raise expectancy: “I can probably complete this now.”

AI can lower cost: “This will take twenty minutes rather than three hours.”

AI can raise utility value: “I can use this skill in more situations because AI fills some gaps.”

AI can increase interest value: “I can experiment rapidly and see results.”

These are genuine motivational advantages.

But AI can also alter value in the other direction. If a task becomes trivial to generate, a person may ask why mastering it matters. If output becomes abundant, status rewards attached to that output may decline. If the person cannot tell whether success reflects their ability, attainment value may weaken. If AI adoption creates pressure, surveillance, job insecurity, or identity threat, psychological cost may rise even while task effort falls.

This helps explain why “AI saves time” is not a complete theory of motivation. Time cost is only one part of the motivational structure.

Human Agency: Motivation Requires More Than Producing an Outcome

Human agency in this article means the person's capacity to govern meaningful parts of goal-directed activity: forming intentions, defining the problem, setting criteria, choosing actions, monitoring consequences, revising strategy, and accepting responsibility.

AI can participate in every one of those steps. That is precisely why the motivational question becomes more important as systems become more capable.

Research on human–automation interaction has long shown that subjective agency can change when control is shared with automated systems. A 2022 review of sense of agency in human–machine interaction describes how automation and assistance can alter the experienced relation between action and outcome Pagliari et al., 2022. Generative AI intensifies this problem because it can contribute not only execution but also language, options, reasons, plans, and evaluation.

The English Psychology Hub treats the broader governance of reasoning in Cognitive Agency in the Artificial Era: Who Governs the Thinking Process?. The boundary is important: that article owns the broader problem of who governs a cognitive trajectory. The present article asks the narrower motivational question: how does that allocation of control affect willingness to begin, invest effort, persist, and experience progress as one's own?

Motivation can remain high with extensive AI assistance when the person still owns the goal and actively governs the process. It can become shallow when activity turns into repeated acceptance of generated outputs whose purpose, criteria, and consequences the person has barely considered.

Four Different Ways AI Can Enter a Motivated Activity

The phrase “using AI” hides psychologically different arrangements. Evidence from education and work becomes easier to interpret when these arrangements are separated.

AI as a barrier reducer

The system removes a cost that is peripheral to the goal: formatting, transcription, basic translation, routine search organization, or repetitive transformation. Motivation may rise because the person can reach the valued part of the task sooner.

AI as a scaffold

The system provides hints, explanations, examples, feedback, questions, or partial support while the person performs the central reasoning or practice. This configuration can support competence when assistance is calibrated and gradually unnecessary for skills the person is trying to acquire.

AI as a collaborator

The person and system alternate contributions. The person may draft first, ask for critique, compare alternatives, revise, and remain responsible for final judgment. Lee and colleagues' 2026 work suggests that this active mode can preserve self-efficacy, ownership, and meaningfulness better than passive copying Lee et al., 2026.

AI as a substitute

The system performs the central task and the person mainly requests, selects, or submits the result. Substitution may be completely rational when the goal is the external outcome rather than skill development. It becomes motivationally costly when the displaced activity was itself the source of mastery, identity, meaning, or independent capability.

None of these modes is universally right. The correct configuration depends on what the person is trying to achieve.

For the distinct question of whether instant answers strengthen or weaken question-asking, exploratory drive, and active learning, see Curiosity in the Age of AI: Learning, Question-Asking, and the Risk of Cognitive Passivity.

Motivation in Learning: The Evidence Is Positive on Average and Highly Context-Dependent

Education currently provides the richest empirical literature on generative AI and motivation.

The newest dedicated meta-analysis, published September 24, 2026, synthesized 42 controlled or quasi-experimental studies and found a positive average effect on learning motivation. Yet heterogeneity was 93.7%, an extremely high value indicating that results differed substantially across studies. The authors therefore caution against treating the pooled average as a universal effect and describe subgroup findings as exploratory because the moderators were study-level characteristics rather than experimentally manipulated variables Fang et al., 2026.

A 2025 systematic review and meta-analysis focused on university students similarly found positive effects on motivation and cognitive, behavioral, and emotional engagement, with variation by subject, learning strategy, and context. It did not find a significant moderation pattern for agentic engagement Xia et al., 2025.

A separate 2025 meta-analysis of university learning outcomes reported positive average effects on affective-motivational outcomes, academic achievement, and higher-order thinking, but no statistically significant average effect on metacognition Chen & Cheung, 2025.

Together, these findings support a careful conclusion: generative AI can improve motivation in learning settings, but motivational improvement does not guarantee deeper learning, stronger metacognition, or durable independent capability.

That distinction is central because AI changes the relationship between experience and evidence. A learner may feel more successful because assistance is effective. The learner may indeed be learning more. The learner may also be seeing the performance of a human–AI system and interpreting it as personal mastery. Those possibilities must be measured separately.

The dedicated English Psychology Hub article Learning in the Artificial Era: AI Scaffolding, Dependence, and Cognitive Agency owns the broader learning question. Here, the key motivational lesson is that the best learning design does not maximize either difficulty or convenience. It aligns assistance with the capability the person is trying to build.

Motivation at Work: Productivity and Psychological Connection Can Diverge

At work, organizations often introduce AI for speed, scale, or output quality. Those outcomes matter, but they do not measure motivation directly.

The 2026 Scientific Reports study by Lee and colleagues provides unusually direct evidence. Participants assigned to passive AI use could produce work with assistance, yet they reported lower independent self-efficacy, psychological ownership, and meaningfulness. Participants who first drafted themselves and then used AI to refine their work retained a stronger psychological connection to the task Lee et al., 2026.

This finding makes a critical distinction visible: productivity can rise while one motivational foundation weakens.

The effect should not be generalized beyond the studied tasks as if all AI use at work has the same consequence. Work differs in complexity, identity relevance, accountability, expertise, team structure, and the degree to which employees choose how AI is used. Yet the study demonstrates a mechanism organizations can no longer ignore. If AI removes the very activities through which workers experience competence and ownership, efficiency can carry a psychological cost.

The wider work context is covered in Work in the Age of AI: Human–AI Collaboration, Job Identity, Well-Being, and Inequality. The motivational question inside that larger domain is whether AI-enabled job design leaves people with meaningful goals, intelligible responsibility, opportunities for mastery, and evidence that their contribution matters.

Why Easy Success Can Become Motivationally Ambiguous

A recurring feature of generative AI is immediate fluency. The system can produce an answer before the person has fully represented the problem.

That can feel motivating at first because progress is fast. It can also create three ambiguities.

Was the task mastered or merely completed?

If success depends heavily on assistance, the person may not know what they can do independently. This is the performance-versus-learning problem Yan et al., 2025.

Was the result mine?

Psychological ownership depends partly on control, intimate knowledge of the work, and personal investment. Passive AI use can weaken that connection even when the output is good Lee et al., 2026.

What should I try next?

Mastery normally produces information about the next attainable challenge. If the system can leap directly to advanced outputs, the person's own competence gradient can become harder to read. They may not know which challenge is appropriately difficult for independent growth.

These ambiguities help explain why high satisfaction with AI-generated results should not automatically be interpreted as strong motivation or development.

The Self-Efficacy Paradox: “I Can Do More” and “I Can Do Less Without It” Can Both Be True

AI can expand effective capability at the level of the combined human–AI system. A person can write in unfamiliar formats, analyze more information, prototype faster, communicate across languages, or attempt technical tasks that previously felt inaccessible.

At the same time, dependence on that system can reduce confidence in unaided performance. Zhang and Xu's study explicitly describes this tension between enhanced self-efficacy and technological dependence in university students Zhang & Xu, 2025. Lee and colleagues provide experimental evidence that passive reliance can lower AI-independent self-efficacy Lee et al., 2026.

There is no contradiction. The two beliefs have different reference points.

A person can reasonably believe, “With AI, I can accomplish more,” while simultaneously believing, “Without AI, I am less capable than I used to feel.”

For motivation, this distinction matters because self-efficacy influences whether a person attempts difficult tasks and persists through setbacks. If confidence becomes entirely contingent on tool availability, motivation may become similarly contingent.

A robust AI-supported workflow therefore preserves opportunities for independent mastery in domains where independent capability remains important.

Overreliance Is Not the Same as Frequent Use

Heavy AI use is not automatically psychological dependence.

A professional may use AI every day while retaining clear goals, domain expertise, verification habits, and the ability to work without it when necessary. Another person may use it less frequently yet depend on it for initiating any difficult task.

The 2026 scoping review on GenAI and learner agency treats dependence and overreliance as patterns involving uncritical uptake, cognitive offloading, weakened judgment, and loss of self-regulation rather than mere frequency Wang et al., 2026. A 2026 mixed-methods study likewise found that overreliance was concentrated in a smaller subgroup and related to self-regulatory processes in complex ways rather than following a simple “more use equals more dependence” rule Galindo-Domínguez et al., 2026.

The relevant questions are functional:

Can the person define a goal before asking the system?
Can they evaluate the answer rather than merely accept it?
Can they explain why a recommendation is appropriate?
Can they notice when the system is wrong or irrelevant?
Can they continue when AI is unavailable?
Are they still learning the capabilities that matter to them?
Do they decide when to delegate, or has delegation become the default?

These questions measure agency more directly than screen time or prompt count.

When AI Is Likely to Support Motivation

Current theory and evidence suggest several conditions under which AI is more likely to support rather than displace motivation.

When it lowers an irrelevant barrier

A person may value a goal but be blocked by formatting, language, routine administration, search organization, or another peripheral cost. Removing that cost can increase the probability of starting and persisting.

When it provides feedback rather than only answers

Feedback can make progress visible and support self-regulation. AI is especially useful when it helps a person compare their current work with criteria, detect gaps, or generate questions for revision.

When challenge remains calibrated

A task that is impossible is demotivating; a task that is trivial can also be disengaging. AI can help adjust challenge by providing hints, partial examples, or graduated support rather than immediately performing the whole task.

When the person retains goal ownership

AI-generated plans are most useful when they serve a goal the person understands and endorses. The person remains able to revise the goal, reject steps, and define what counts as success.

When assistance produces mastery evidence

If AI support helps the person perform a process they can later reproduce more independently, it can strengthen competence and self-efficacy.

When collaboration preserves contribution

In work settings, active collaboration that keeps the person's own draft, judgment, or decision central may preserve ownership and meaning better than passive copying Lee et al., 2026.

When AI Is More Likely to Weaken Motivation

Risk rises when AI changes the motivational structure of the task in the opposite direction.

When the outcome arrives before the goal is mentally represented

If a person asks for “something good” and immediately receives a polished artifact, they can skip the process of deciding what they are trying to accomplish. Repetition of that pattern can make goal formation itself a delegated activity.

When the person cannot attribute success

Strong output with ambiguous authorship can fail to build independent self-efficacy. The person knows the result is good but cannot locate their own competence within it.

When the task was supposed to train the person

Substitution is particularly risky when the purpose is learning. If the system performs the exact cognitive operation the learner needs to practice, immediate performance can conceal a developmental loss.

When AI removes every useful difficulty

Not every difficulty should remain. But learning research shows that some effortful conditions improve retention and transfer de Bruin et al., 2023. If AI always eliminates retrieval, generation, explanation, error correction, and sustained attention, it can remove the mechanisms that build durable skill.

When external pressure dominates

Organizations and schools can turn AI from a source of autonomy into a requirement: use this system, meet a higher quota, respond faster, produce more. Under those conditions, technology that technically expands options can psychologically increase control.

When social motivation is displaced

If AI replaces human feedback, collaboration, recognition, or mentorship rather than complementing it, relatedness may weaken even when task support improves. Current classroom evidence is not sufficient to generalize this effect, but it supports treating social motivation as a separate outcome rather than assuming it will follow from better individual assistance Schweder et al., 2026.

A Practical Way to Use AI Without Outsourcing Motivation

There is no universal percentage of a task that should remain “human.” The useful unit is function, not percentage.

Define the goal before asking for the path

Write down what you are trying to achieve and why it matters. Then use AI to improve the route. This keeps goal ownership visible.

Decide whether the current objective is output or capability

If the objective is output, delegation may be efficient. If the objective is capability, preserve the cognitive operation you are trying to learn. Do not ask AI to perform the very skill you intend to acquire.

Make an initial attempt when mastery matters

A first attempt generates diagnostic information. It reveals what you know, where you are stuck, and what kind of help you actually need. It also gives later improvement something to build on.

Ask for feedback, alternatives, and questions

Instead of requesting the final answer every time, ask the system to critique your reasoning, identify missing evidence, pose questions, compare approaches, or explain an error. These uses keep the person cognitively active.

Separate assisted performance from independent performance

Periodically test what you can do without AI. The point is not ritual purity. It is calibration. If independent capability matters, you need evidence about it.

Keep the final criterion human-readable

Know why the final output is acceptable. “The AI said so” is not a criterion. A good workflow leaves the person able to state the standards used to evaluate the result.

Preserve some tasks that produce mastery

If writing an argument, solving a problem, remembering a concept, or making a judgment is central to your competence, retain regular opportunities to perform that operation directly.

Watch motivation over time, not only satisfaction in the moment

AI can make a task immediately more enjoyable because frustration drops. That is useful. The longer question is whether you are becoming more willing and able to engage with the domain, or increasingly unwilling to begin without assistance.

For Students and Learners

Students should distinguish three questions that are often collapsed into one.

First: Did I complete the assignment?

Second: Did I understand the material?

Third: Could I use the knowledge later without the same support?

Generative AI can improve the first answer without guaranteeing the second or third. The latest meta-analytic evidence suggests genuine average benefits for motivation and several learning outcomes, but also substantial heterogeneity and unresolved questions about metacognition and dependence Chen & Cheung, 2025 Fang et al., 2026.

A strong learning workflow uses AI in ways that create more high-quality practice, not merely fewer opportunities to struggle. Ask for explanations after an attempt. Generate additional examples after you have classified the first ones yourself. Request feedback on a draft rather than replacing drafting altogether. Use AI to create quizzes, then retrieve answers before looking at the explanations.

The best indicator is transfer: can you solve a new problem, explain the concept, or make a judgment when the prompt is different and assistance is reduced?

For Educators

Educators should not evaluate AI integration only by asking whether students like it or whether immediate performance improves.

Motivation has quality. An intervention can increase enjoyment while leaving self-regulation unchanged. It can increase perceived competence while making independent competence harder to evaluate. It can raise autonomous motivation for some students while having little effect on others.

The 2026 lower-secondary study by Schweder and colleagues is especially informative because it found different motivational profiles rather than a uniform shift across students Schweder et al., 2026. The September 2026 meta-analysis likewise found large heterogeneity across studies Fang et al., 2026.

Educational design should therefore specify the role of AI. Is it explaining, hinting, critiquing, generating practice, modeling, checking, or doing? Assessment should also distinguish AI-supported production from independent knowledge when independent knowledge is the target.

The broader educational implications are developed in Education in the Age of AI: Learning, Motivation, Assessment, and Cognitive Development.

For Workers and Professionals

For professionals, the central danger is not “using AI too much” in the abstract. It is losing contact with the functions that sustain expertise, judgment, and ownership.

Routine automation can be beneficial. But if AI gradually takes over problem framing, first drafts, evidence selection, evaluation, and final recommendation, the human role may shrink to acceptance. That configuration can preserve output while weakening the person's evidence of competence.

A better professional workflow keeps responsibility visible. Define the objective. Decide what evidence matters. Use AI for search support, alternatives, critique, transformation, or acceleration. Verify consequential claims. Retain final judgment in decisions for which you remain accountable.

The same principle explains the experimental difference between passive copying and active collaboration in the Lee study Lee et al., 2026.

For Managers and Organizations

Organizations can unintentionally demotivate workers by treating AI adoption as a pure productivity program.

If every gain in speed becomes a higher output quota, employees may experience AI as intensified control rather than support. If workers lose authorship over outputs for which they remain accountable, psychological ownership can decline. If junior employees no longer perform tasks through which expertise was historically acquired, the organization can improve short-term throughput while weakening its future skill pipeline.

AI implementation should therefore be evaluated on at least four outcomes: performance, learning, independent capability, and psychological connection to the work.

Managers should also distinguish AI self-efficacy from professional self-efficacy. Training people to prompt a system is useful, but it does not automatically preserve confidence in the underlying professional task.

Motivation, Self-Worth, and Social Comparison

AI adds a new comparison target. A person can compare themselves not only with peers but with a system that can generate fluent output in seconds.

That comparison can motivate exploration: “I can use this to do more.” It can also destabilize ability beliefs: “If the system can do this instantly, what does my competence mean?”

These questions belong partly to self-worth rather than motivation. The English Psychology Hub therefore treats ability, comparison, and status in a separate canonical owner: Self-Worth in the Artificial Era: Ability, Comparison, Status, and AI. The present article keeps the boundary narrower: comparison matters here when it changes expectancy, persistence, goal choice, or willingness to invest effort.

Motivation and Meaning Are Related but Not Identical

A meaningful activity can motivate effort, and effort can contribute to meaning, but motivation and meaning should not be collapsed into one construct.

Recent review work argues that AI may create a paradox in which people experience less meaning through reduced effort, self-efficacy, mattering, or cultural stability while simultaneously needing more meaning as technological change challenges familiar sources of human significance. The authors describe this as a conceptual account, not proof that AI inevitably causes meaninglessness Mead et al., 2026.

The English Psychology Hub develops this broader question in Meaning in the Artificial Era: Work, Effort, Selfhood, and Human Significance. For motivation, the narrower consequence is that people may become less willing to invest effort when they no longer see why the effort matters, even if the external goal remains available.

What the Evidence Does Not Yet Establish

The research base is expanding quickly, but several conclusions would currently go beyond the evidence.

We do not yet have strong longitudinal evidence showing that ordinary long-term use of general-purpose generative AI inevitably reduces human motivation. Many studies are short-term, educational, cross-sectional, quasi-experimental, or focused on specific tasks.

We do not have a universal threshold at which AI assistance becomes dependence. Functional loss of self-regulation matters more than a simple usage count.

We cannot assume that findings from students transfer directly to experienced professionals, or that results from writing transfer to coding, therapy, management, caregiving, or creative practice.

We cannot infer independent competence from AI-assisted output quality.

We cannot infer low motivation from reduced effort. AI may remove unnecessary costs while motivation remains high.

We also cannot infer healthy motivation from high engagement alone. A person may spend a great deal of time with an AI system because the interaction is rewarding without developing a durable goal, skill, or autonomous reason for action.

These limitations are not a reason to avoid conclusions. They define the conclusions the evidence currently supports: AI's motivational effects are conditional, mechanism-specific, and strongly shaped by task design and mode of use.

A Research Agenda for Motivation in the Age of AI

The most important next studies should move beyond asking whether people “used AI” and measure how cognitive and motivational labor was divided.

Researchers need to distinguish at least assisted performance, independent performance, AI self-efficacy, task self-efficacy, autonomous motivation, controlled motivation, goal ownership, persistence, psychological ownership, and transfer.

Longitudinal studies are especially important. A tool can increase motivation during the first weeks because it is novel, helpful, or relieving. The more important question is what happens after months or years: whether people attempt harder goals, learn more efficiently, preserve independent capability, and remain willing to engage when assistance is unavailable.

Research also needs better task-level descriptions. “ChatGPT use” can mean asking for a hint, generating a complete answer, receiving Socratic questions, editing a draft, brainstorming, tutoring, translating, planning, coding, or emotional support. Treating these as a single exposure obscures the mechanisms that matter.

Finally, motivation should be studied across the life course and outside education. Evidence on adolescents, mid-career workers, older adults, entrepreneurs, creators, caregivers, and unemployed people adapting to AI-mediated labor remains much thinner than the higher-education literature.

Frequently Asked Questions

Does AI make people lazy?

“Lazy” is too imprecise to describe the evidence. AI can reduce effort because a task becomes more efficient, and that can be beneficial. It can also encourage offloading of cognitive work that would otherwise support learning, judgment, or mastery. The relevant question is what kind of effort disappeared and whether the person's capability, agency, or motivation changed.

Can AI improve motivation?

Yes, in some contexts. Recent meta-analyses in education find positive average effects on learning motivation and engagement, but effects vary substantially across studies and settings Fang et al., 2026 Xia et al., 2025.

Can AI reduce self-efficacy?

It can under some patterns of use. Passive reliance in a 2026 work experiment reduced confidence in completing the task without AI, while active collaboration largely preserved that confidence Lee et al., 2026. Other studies show that AI can also increase confidence and perceived efficiency, which is why the reference point—using AI versus acting independently—must be specified.

Is less effort always bad for motivation?

No. Removing pointless effort can increase motivation by reducing cost. Effort becomes important when it contributes to learning, mastery, ownership, or meaning. Psychology supports selective preservation of useful effort rather than maximum difficulty.

Should I always try a task without AI first?

Not always. A first attempt is especially useful when your goal is learning, diagnosis of your current skill, or preservation of independent competence. If the task is routine and your goal is simply an accurate external outcome, direct delegation may be efficient.

Does AI undermine intrinsic motivation?

Current evidence does not support a universal claim. AI can increase interest and perceived competence in some settings, while controlling use, passive substitution, or reduced ownership can work in the opposite direction. Self-determination theory predicts that the effect will depend heavily on autonomy, competence, and relatedness.

What is the difference between AI self-efficacy and task self-efficacy?

AI self-efficacy is confidence in one's ability to use AI effectively. Task self-efficacy is confidence in one's ability to perform the underlying task. They can move together, but they do not have to. A person can become better at directing AI while feeling less able to perform independently.

Can AI help with goals?

Yes. It can clarify, decompose, schedule, compare, and monitor goals. Motivation is more likely to remain autonomous when the person still understands and endorses the higher-order goal and can revise the AI-generated plan.

How can I tell whether I am becoming overreliant on AI?

Look at function rather than frequency. Warning signs include inability to start without AI, uncritical acceptance of outputs, reduced ability to explain or verify results, avoidance of independent practice in skills you want to retain, and uncertainty about your own standards or goals.

What does human agency mean in AI-assisted motivation?

It means retaining meaningful control over goal-directed activity: forming intentions, defining criteria, choosing actions, evaluating outputs, revising course, and accepting responsibility. AI can support those functions without needing to replace them.

Is motivation in the Age of AI mainly an education problem?

No. Education currently has the largest research base, but the same mechanisms matter in work, creativity, entrepreneurship, self-directed learning, and everyday goal pursuit. Evidence should still be transferred cautiously across domains because tasks and incentives differ.

What is the difference between the Age of AI and the Artificial Era?

Age of AI is broad contemporary language for a period shaped by AI technologies. Artificial Era is Angela Bogdanova's formalized Aisentica category for a historical-philosophical condition in which Artificial becomes an independent non-biological order beside Homo. They overlap in contemporary context but are not synonyms Bogdanova, 2026.

Conclusion: Motivation Depends on What AI Removes, What It Preserves, and Who Still Governs the Goal

AI changes motivation because it changes the structure of action.

It changes how hard a task feels, how quickly progress appears, whether success seems attainable, where feedback comes from, whose competence produced the result, who selected the goal, how much effort remains, and whether the person can still see a causal line between intention and outcome.

That is why the same technology can motivate one person and demotivate another. The same person can even experience both effects at once: greater willingness to attempt a task with AI and lower confidence about performing it without AI.

The evidence available by September 2026 supports neither a simple optimism nor a simple decline narrative. Generative AI shows positive average effects on motivation in multiple educational syntheses, yet those effects are highly heterogeneous. Work experiments show that active collaboration can preserve self-efficacy and meaning more successfully than passive reliance. Research on cognitive offloading and learner agency shows both supportive and erosive patterns depending on how assistance is embedded.

The durable psychological principle is therefore functional. Remove effort that is merely a barrier. Preserve effort that builds the capability you value. Use AI to make goals clearer without letting it decide what matters. Let assistance increase competence without confusing system performance with personal mastery. Keep enough independent practice to know what you can do. Retain final judgment where responsibility remains yours.

Motivation in the Age of AI is ultimately a problem of human agency: not whether people will continue to act when machines can do more, but whether people remain able to choose their goals, recognize their own competence, invest effort where it matters, and understand why the action is worth taking.

Cognitive Agency in the Artificial Era: Who Governs the Thinking Process?

Learning in the Artificial Era: AI Scaffolding, Dependence, and Cognitive Agency

Meaning in the Artificial Era: Work, Effort, Selfhood, and Human Significance

Self-Worth in the Artificial Era: Ability, Comparison, Status, and AI

Work in the Age of AI: Human–AI Collaboration, Job Identity, Well-Being, and Inequality

Curiosity in the Age of AI: Learning, Question-Asking, and the Risk of Cognitive Passivity

Human Needs in the Age of AI: Autonomy, Competence, Relatedness, and Meaning

Psychological Encyclopedia