What are the cognitive and motivational mechanisms by which using AI to copy answers reduces college students' ability t

What copying answers from AI does in practice

Students who over-rely on generative AI tend to hand the assignment to AI without changing it and then adopt the AI-generated output uncritically as their own solution [35]. They also tend to prioritize immediate AI help over independent problem-solving, and this behavior comes with reduced cognitive effort [35][38]. Students may view GenAI as an inevitable part of higher education while still recognizing overt cheating [66]. People without research backgrounds also tend to over-rely on AI to answer questions, which matters for students still developing those skills [30].

Cognitive mechanisms

Cognitive offloading. Using AI encourages cognitive offloading, which means handing a mental task to an outside tool [11][20]. Research suggests this process can reduce critical thinking [1], and when it is overused it risks eroding critical thinking [12]. AI users may spend less time on deep, reflective thinking and prefer quick AI-generated solutions [15]. Education bodies warn that over-reliance on generative AI can lead to cognitive offloading and undermine critical thinking [36].

Bypassing recall and problem-solving. Copying answers skips the mental work that builds skill. AI tools reduce the opportunity for active recall and problem-solving, which are essential parts of cognitive development [2]. Defaulting to AI responses rather than reasoning through problems can weaken cognitive flexibility and information evaluation skills [17]. Overuse of AI answers reduces the germane load, meaning the mental effort needed for deep learning and higher-order thinking [6], and can impair independent critical thinking [7]. AI can turn students into passive consumers rather than active thinkers [13] and can discourage them from engaging in problem-solving and analytical exercises [16]. Over-reliance also risks diminishing information retrieval skills, deep knowledge exploration, critical thinking, and creative problem-solving [37], and it can undermine the foundational cognitive skills needed for long-term academic growth [41].

Mental effort, attention, and memory. Excessive reliance on AI may reduce cognitive engagement and long-term retention [3]. Prolonged AI exposure was linked to memory decline in a study of undergraduates [4]. Students may passively accept AI-provided information without critical scrutiny [5]. Over-reliance lowers the quality and extent of mental work and problem-solving ability [10], and tools that reduce mental load can harm thinking [23]. Reported effects include shorter attention span, weaker concentration, more impulsivity, and less time for critical thinking [25]. People who used large language models to complete tasks underperformed at neural, linguistic, and behavioral levels compared with people who used their own brains [26].

Metacognition and verification. Metacognition means thinking about your own thought processes [21]. Using AI without active metacognition is described as fundamentally dangerous [22]. Trust in AI-generated content is linked to reduced independent verification of information [18], and some users report rarely reflecting on the biases behind AI recommendations and trusting outputs outright [31]. Healthy learning includes a calibration step: feedback after attempting to recall helps students judge what they know and do not know, but copying answers bypasses that retrieval-and-feedback step [69][70][76].

Correlational evidence. Studies find a negative correlation between frequent AI tool use and critical-thinking ability [14][27]. Younger individuals tend to show stronger dependence on AI and score lower on critical-thinking assessments than older participants [19]. This fits the broader finding that reliant behavior diminishes active cognitive participation and inhibits learning ability [46].

Motivational mechanisms

Intrinsic motivation fades. When AI does the thinking, students may lose motivation, engagement, and the intrinsic motivation to learn and solve problems independently [9]. Over-reliance promotes superficial learning because it prioritizes rapid access to information over meaningful understanding, which reduces motivation for active knowledge exploration [40].

Autonomy, competence, and relatedness. Self-determination theory says motivation is strongest when three basic needs are met: autonomy, competence, and relatedness [42][49][58]. Autonomy means feeling in control of your own behavior [59], and autonomous people are more likely to persist in the face of challenges, while people who feel controlled are less motivated [65]. Excessive dependence on AI can compromise autonomy [8]. Higher perceived autonomy is linked to activating metacognitive abilities and critically evaluating AI outputs rather than passively receiving them [48]. In education, this need-based framework has been used to promote student engagement and motivation [62].

Low competence pushes students toward copying. Competence means knowing you have the skills and knowledge to handle a novel challenge [60]. Students with low perceived competence may doubt their own abilities, avoid cognitive challenges, and rely directly on AI to complete assignments [44]. Competence satisfaction is especially important in AI-enriched learning environments [53][54]. Perceived competence directly promotes deep processing and reduces excessive reliance [47]. When students feel autonomous, competent, and connected, their intrinsic motivation is strengthened and they engage in deeper cognitive processing rather than superficial over-reliance [43].

Guilt and expectations instead of enjoyment. Many students learn and use AI mainly out of guilt or shame rather than personal enjoyment [50]. Introjected regulation, meaning acting from guilt, shame, or perceived expectations, was the central driver of students’ AI learning, while intrinsic motivation was less central [50][51]. Students often engaged because they felt it was expected of them by others [52], and their motivation was fueled largely by guilt avoidance [55]. This is a controlled, low-autonomy form of motivation [65].

Dependence grows and motivation shifts over time. Initial excitement about AI can fade, and some students may become distracted by AI over time [56]. Heavy AI use can work against skill development, especially before a skill has been adequately developed [24]. People report worrying that relying on AI means they are not really learning and would not know how to solve problems without it [28]. Others report that the more they use AI, the less they feel the need to solve problems on their own [29]. One useful framing from the evidence is that AI should be supplemental to thinking, not a replacement [32]. An analogy used in the evidence is navigation apps: they give confidence to visit new places but can inhibit the sense of direction and create unintended dependencies [63]. Applied to coursework, the concern is that AI-generated work may not be committed to memory and may weaken independent critical evaluation [64]. Students’ attitudes toward AI also affect their AI-assisted creativity through their motivation to use AI [34].

Why this is the opposite of how learning works

Learning science shows that recalling information from memory, not rereading or being handed answers, is what builds durable learning [67][77]. Exams require retrieving information from memory without notes or slides [68]. The act of being quizzed itself helps students learn better [72], because pulling concepts out of the brain is more effective than having concepts presented [73]. Flashcards only help when students try to recall the answer before seeing it [74], and the struggle to recall strengthens long-term learning [75]. Checking course materials works best after an attempted recall, as in a copy-cover-and-check method [69][70].

Research comparing retrieval with worked examples, meaning seeing a problem and an explanation of the answer, found that retrieval practice outperformed worked examples in every condition when students had been taught the material first [78][80][81]. Worked examples can help in some narrower conditions, and the best strategy depends on the learning context and goals [82][86]. But retrieval practice takes longer and is more effortful than being shown solutions [84], which is exactly why copying AI answers can feel easier even though it skips the retrieval step that supports independent learning [71][74].

The combined effect

Put together, the cognitive and motivational mechanisms point in the same direction. Copying answers offloads mental work, skips active recall, weakens attention and memory, and bypasses the feedback that calibrates self-judgment [5][10][23][76]. At the same time, it lowers intrinsic motivation, undercuts autonomy and competence, and shifts students toward guilt- and expectation-driven learning [9][44][50][65]. The result is that students become more dependent on AI and less able, and less motivated, to learn on their own [7][28][41].