Cognitive Miser Theory as a Lens for Rethinking Cognitive Effort During Difficult and Complex Academic Tasks in AI-Supported Self-Regulated Learning

Cognitive Concerns on AI Use

Artificial intelligence (AI), with the potential to improve cognitive processes, reshape educational methodologies, and provide adaptable personalized learning experiences, brings about concerns regarding its impact on the use of cognitive resources and learning efficacy, as well as social and ethical matters (Dergaa et al., 2024; Gkintoni et al., 2025; Han et al., 2024; Katsantonis & Katsantonis, 2024; Karaoglan Yilmaz & Yilmaz, 2025). In the context of retention, problem-solving, critical thinking, and creativity skills, the use of AI as a facilitator that supports SRL and meaningful learning is essential for cognitive development (Jose et al., 2025). Therefore, AI in educational contexts can act as a tool, agent, or method for externalizing, internalizing, or expanding human cognition. While externalization of cognition is the definition and modeling of tasks with an AI tool, in internalization processes, AI can be used to help change thought representations (Cukurova, 2025). In expanding human cognition, AI are being used in a hybrid approach based on human-AI interaction. Therefore learning, cognition, and AI paradigms should be examined within the context of cognitive load theories of learning (Mayer, 2013; Sweller, 2020) and are related to the effective management of cognitive resources (Jose et al., 2025).

By offering interactive, human-centered, and personalized experiences, AI supports cognitive processes and presents significant opportunities for focusing attention on information, retaining it in working memory, and integrating new information into long-term memory as prior knowledge (Cukurova et al., 2020; Giannakos et al., 2025; Sánchez Vera, 2024). In particular, AI has strong potential for reducing perceived cognitive load by enabling the offloading of working memory capacity to more complex cognitive tasks (AlShaikh et al., 2024; Ebadi & Amini, 2024; Feng, 2025; Gerlich, 2025; Meng et al., 2025; Singha et al., 2024; Yin et al., 2024). However, alongside compelling examples of AI reducing perceived cognitive load, there are concerns that it may diminish cognitive capacity due to excessive cognitive overload or cognitive offloading. In critical AI use, where expertise and metacognitive awareness are high, perceived cognitive load increases over time (Schulze et al., 2025). Conversely, in situations where expertise, metacognitive awareness, and a critical approach decline while cognitive offloading increases uncontrollably, AI use impairs the utilization of cognitive capacity (Kosmyna et al., 2025; Yiğit, 2025).

Studies examining situations in which students feel cognitively inadequate or exhibit automatic AI usage behavior when completing complex tasks indicate that such behaviors undermine metacognitive skills such as critical thinking and problem-solving (Jose el al., 2025; Kosmyna el al., 2025; Han et al., 2024; Schulze el al., 2025; Singha el al., 2024; Tsakeni el al., 2025; Zhai el al., 2024). Similarly, when AI is used automatically in time-pressured tasks, students tend to avoid using their cognitive resources, opt for shortcuts, rely on AI, become impatient, and experience a decline in motivation (Li et al., 2026; Kim et al., 2026). However, Li and colleagues (2026), who note that perceived cognitive load increases as task complexity rises, point out that students turn to AI to reduce this load and that this behavior leads them toward an AI dependency that distances them from critical thinking in order to reduce cognitive effort (e.g., cognitive offloading). The replacement of critical thinking with automated, intuitive use leads to cognitive overload and a decline in learner autonomy (Tsakeni et al., 2025). This phenomenon, also known as the “Google effect,” indicates a decline in cognitive resilience due to the loss of practice in tackling challenging tasks; it is likened to atrophy, similar to how a muscle wastes away when not used (Azarov & Gorokov, 2025; Khan & Suhluli, 2025; Gerlich, 2025; Strati, 2026).

The Relation of Cognitive Miser Theory and Complex/Difficult Tasks

The Cognitive Miser Theory (CMT) posits that individuals tend to conserve cognitive resources by processing only the amount of information they perceive as necessary to form an opinion or make a decision, rather than engaging in exhaustive analysis. This tendency often results in decisions made with insufficient information, as individuals are motivated to minimize mental effort unless the perceived benefit justifies greater cognitive investment (Todaro et al., 2023). People explicitly and implicitly seek to minimise cognitive effort in decision-making, interpreting any excessive use of cognitive resources as a cost. They typically expend only the effort required to reach a satisfactory, rather than optimal, decision (Todaro et al., 2023).

The theory is closely linked to dual-process models, which distinguish between fast, heuristic (Type 1) and slow, systematic (Type 2) processing. Cognitive misers default to heuristic processing whenever possible, relying on mental shortcuts, emotions, and surface-level information to make quick judgments (Toplak et al., 2014). There is a recognised trade-off between the effort invested in decision-making and the accuracy of the outcome. As cognitive effort decreases, decision accuracy may also decline, but individuals aim to balance effort and accuracy to achieve an adequate result without unnecessary cognitive expenditure. Also, due to high demands reduce efficiency and decision quality (Fan et al., 2025), perceived cognitive load (the total mental effort required) plays a pivotal role.

AI systems are designed to address cognitive challenges similar to those solved by humans, converting cognitive capabilities into computational formulas to solve problems and make decisions. The ultimate goal is to achieve human-level intelligence in machines, with applications spanning various domains such as education, business, and healthcare. In summary, the CMT provides a framework for understanding how doctoral students might interact with AI, often using them to minimise cognitive effort in complex academic tasks. This is because the limits of cognitive resources are closely related to not only the design of learning materials but also to the cognitive processes of individuals (Mayer, 2013; Mutlu-Bayraktar, 2024). This interaction is shaped by cognitive load considerations, trust in AI, and ethical decision-making, highlighting the intertwined roles of cognitive psychology and AI in academic research environments (Todaro et al., 2023). Such interaction shaped by cognitive resources can be limited by complex and difficult tasks, which include understanding, thinking, relating, planning or conducting a study, academic writing, making a holistic literature review, time and resource management, analyzing, synthesizing, and evaluating the information, etc. For the judicious management of cognitive resources, learning experiences should be designed to account for tasks that demand high cognitive involvement and complexity.

Examining AI in Education Within The Critical Lens of Cognitive Miser Theory

Cognitive miserliness can be a risk to meaningful learning and the cognitive processing stages (selection, organization, integration) (Mayer, 2013) in the realization of deeper learning. Although perceived cognitive load is predicted to decrease in AI-supported applied learning practices (Feng, 2025), the use of AI in active cognitive learning processes has been observed to increase perceived cognitive load and negatively affect the use of cognitive resources (Jose, et al. 2025; Kosmyna et al., 2025). This is thought to increase preferences for easy and unquestioning access to information, particularly in line with current AI-based learning approaches, thus jeopardizing cognitive development in academic tasks (Flores et al., 2025; Jose et al., 2025; Zhai et al., 2024). Therefore, for the effective and efficient use of AI in education, it is important to adopt an approach that focuses on cognitive development in AI-supported learning processes and ensures the efficient use of cognitive resources. It is important to preserve and develop skills such as retention, problem solving, critical thinking and creativity, especially in difficult and complex tasks where AI is used, by taking into account intuitive approaches that are fast, automatic, and performed with low cognitive effort.

Research Context

The data of this study were collected from two universities that offer educational opportunities through educational technology and distance education programs. Since there were only three doctoral-level educational technology programs in Türkiye, the target audience was expanded to include doctoral students in distance education programs. In accordance with the targeted criteria, doctoral students at two institutions were reached. Before participants were included in the study, the ethics committee of the university to which the researchers are affiliated reviewed and approved the study’s appropriateness. The necessary permissions to conduct the study were obtained, and the procedures in the study were carried out in accordance with the ethical standards of both institutions. The participants, who are pursuing degrees at two different universities in Türkiye, representing diverse educational backgrounds, were doctoral students studying educational technology and distance education. The participants included students who are currently in the thesis or course phase of their doctoral studies, use AI in their academic work on a daily basis, and have a long-time interaction and experience working with various AI tools. In this study, in which participation was entirely voluntary, no comparisons were made between participants’ individual or organizational uses of AI. Although this case study was conducted on two seperate organizations, the study sought to gain an in-depth understanding of the participants’ shared experiences with AI use. Because the participants comprised a diverse range of adult learners actively engaged with AI-enhanced learning resources, this group provided a methodologically relevant context for understanding the reasons for using AI in complex and difficult academic tasks. Data were collected through semi-structured interviews, allowing for the participation of individuals within a context relevant to the case identified within the study.

The participant group included in the study through purposive sampling (Table 1) consists of individuals (n = 7) who are continuing their doctoral studies in educational technology or distance education. The participant group was recruited via social media, prioritizing accessibility and ease of reach to maximize accessibility. Voluntary participation was fundamental to the inclusion of all participants. The criteria for including participants in the study were “having a postgraduate education at a state or private university in Türkiye”, “continuing doctoral studies in educational technology or distance education”, “frequently using AI in academic and independent learning processes”, and “voluntarily participating in research”.

The researchers, who are a full-time faculty member and a doctoral student at one of the two universities where the study was conducted, clearly disclosed their positions at the relevant institution prior to the semi-structured interviews to ensure the study’s transparency and reflexivity. Since participants were contacted through informal WhatsApp groups, necessary precautions were taken to minimize any biases that might arise from institutional ties. It was guaranteed that the study would be conducted based on the students’ voluntary participation, that students could withdraw from the study at any time, that they would not be subject to any grading or scoring system, and that the interviews would be free from academic pressure. In addition, it was clearly stated that the confidentiality of the data collected would be protected and that personal information would be concealed using pseudonyms.

Data regarding the defined situation were collected through semi-structured interviews with the participants between May and June 2025. The interview questions were prepared based on the Cognitive Miser Theory and expert opinions. Participants were informed about the research, and their approval for recording the interviews and for their voluntary participation was obtained. Each interview, facilitated through video conferencing, had an approximate duration of 30 minutes. The records obtained as a result of the interviews were transcribed.

Data Analysis

Thematic analysis of data obtained through semi-structured interviews was conducted using an open coding approach (Braun & Clarke, 2019). The codes and themes were generated inductively from participants’ accounts rather than being predetermined by the theory. Thematic analysis, which allows for an inductive approach to examine data obtained from different types of sources related to a situation, enables in-depth examination, classification, evaluation, correlation, description, and interpretation of different dimensions of situations (Fraenkel et al., 2023). In this research, the data obtained and transcribed were first read from beginning to end, and then, considering the general structure of these data, codes, subthemes, and themes were extracted using an inductive approach. The findings obtained were described and interpreted to answer the research question. The qualitative data obtained in this study were analyzed using the thematic analysis method suggested by Braun and Clarke (2019). Thematic analysis is defined as the process of identifying, analyzing, and reporting meaningful patterns within data and offers a flexible yet systematic approach to qualitative research. This study adopted an inductive approach to develop meaningful themes that directly address the research questions.

During the data analysis process, the data obtained from interview transcripts were first read repeatedly to gain awareness of the overall structure. Qualitative data analysis tools were not employed to achieve a comprehensive understanding of the situation under review. Instead, the analyses were conducted manually by the researchers. Meaningful data fragments were then extracted, and concise, descriptive codes based directly on participant discourse were developed. Similar or related codes were brought together to form meaningful subthemes (categories), which were then grouped under a more abstract theme. The developed theme structure was reviewed and overlapping or weak areas were reorganized.

Trustworthiness

Data obtained from participants during the interviews was verified through member checking (Fraenkel et al., 2023). For member checking, participants were first asked to confirm their responses during the interview in real time. Subsequently, the transcripts prepared after the interviews were presented to the participants for review, and member checking was thus conducted in two stages. Following the analysis phase, participants were asked to evaluate the codes and themes derived from the approved raw data. To ensure credibility, member checking was conducted by inviting participants to interpret and verify the analysis report (Creswell & Creswell, 2018; Tisdell et al., 2025).

Since the data were collected through semi-structured interviews, methodological triangulation was not claimed. Instead, credibility was supported through member checking, negative case analysis, theory triangulation, researcher cross-checking, and expert review. First, the fundamental perspectives of Cognitive Miser Theory were examined to establish a consistent rationale for the creation of codes and themes (Creswell & Poth, 2016; Miles & Huberman, 1994). Then, the analysis was conducted thematically, aligning with the Cognitive Miser Theory. As a third method to strengthen the transferability, negative or inconsistent information was also evaluated, aiming to provide a comprehensive and diversified depiction of the described situation (Miles & Huberman, 1994). The analysis was conducted thematically by the researchers. To ensure the dependability of the findings, the data were compared and interpreted using a codebook (Creswell & Creswell, 2018). The developed codes and themes were independently compared and cross-checked by the researchers (Fraenkel et al., 2023; Tisdell et al., 2025). In the final phase, a field expert confirmed the dependability of the results.

The Role of AI in Academic Decision-Making

The findings obtained in the context of the theme titled “the supportive role in decision-making,” in which the students’ opinions about their thinking processes in difficult and complex academic tasks were taken, were coded under four headings: (a) cognitive support, (b) social and emotional support, (c) validating self-efficacy and self-competence, and (d) providing time and source management.

Students stated that in difficult and complex tasks, they first try to understand and stage the process with their own cognitive efforts and then use AI as an assistant that provides emotional support in managing resources, stress, and time. Although AI is a guide or a consulted information provider, students emphasized that they first made a cognitive effort to make sense of the process. In line with these findings examined under the subtheme of “cognitive support,” it can be stated that students receive assistance from AI in these processes where they exert intense cognitive effort, supporting their internal motivation in both academic and emotional dimensions. In this regard, students described themselves as more comfortable, less stressed, and more confident when using AI. Therefore, students are observed to seek support for structuring the learning process as a result of cognitive effort during the thought processes of AI use. Furthermore, using AI on difficult and complex tasks appears to have a positive emotional impact on students’ thought processes.

Under the subtheme of “social and emotional support,” which includes the emotions underlying the intellectual processes and initial actions taken as a result of AI interaction, students expressed that there were both academic and emotional expectations in their orientation to use AI.

Among the findings examined under the subtheme of “validation self-efficacy and self-competence” are that students’ use of AI in difficult and complex tasks does not affect their feelings of competence and control on familiar topics, but they tend to lean more towards AI competence and control when they are on a new and unfamiliar topic or when AI outputs are used directly.

Findings regarding the perceptions of competence and control in using AI in difficult and complex tasks reveal that another portion of students prioritize their own competence and use AI to reinforce these competencies and to manage resources and time.

In situations where tasks are complex and time-constrained, students report heightened anxiety, suggesting that AI use under temporal pressure may inadvertently reduce engagement with their own metacognitive competencies such as planning and monitoring. These findings emphasize that while AI can enhance control and efficiency, careful scaffolding and training are essential to prevent over-dependence and ensure meaningful learning.

In the “providing time and source management” theme, it was found that motivation, productivity, resources, stress and time management are among the main reasons why students turn to AI technologies when faced with time pressure or an unknown academic task. As can be seen from the findings, the relevant situations appear to involve practical rather than cognitive effort.

The Cognitive Processes

The findings compiled in the theme examining the views on the cognitive effects of using AI in difficult and complex academic tasks were coded under four headings: (a) changes in cognitive effort and perceived cognitive load, (b) affecting creative cognition, (c) original thinking processes, and (d) confronting intellectual dilemmas. Findings regarding the impact of AI use on “changes in cognitive effort and perceived cognitive load” indicate that students’ direct access to information from AI reduces cognitive effort.

Other findings regarding the effects of AI use on difficult and complex academic tasks on cognitive effort and load indicate that cognitive effort, and therefore perceived cognitive load, increases when the information presented by AI is restructured through students’ evaluations and questioning. In this regard, students are observed to expend a high level of cognitive effort in the stages of compiling commands given to AI, arriving at the correct conclusion, and compiling the obtained conclusion.

When the cognitive effects of students’ AI use are examined under the heading of “affecting creative cognition”, it is seen that some students reported feelings of anxiety, intimidation, and insecurity, particularly when comparing their own ideas with AI-generated outputs, others experienced enhanced productivity and curiosity through critical and inquisitive engagement.

The generated codes indicate that some students’ productivity and creativity were supported by their inquisitive and critical use of AI, which positively impacted their emotions.

In the subtheme examining the cognitive impact of AI use on difficult and complex tasks, the codes obtained under the heading “original thinking processes” indicate that students’ original thinking processes are not negatively affected during AI use; on the contrary, they emphasize the importance of individual contribution and evaluation.

The codes examined under the heading “confronting intellectual dilemmas” indicate that students’ cognitive awareness is high during AI use; accordingly, they do not experience any dilemmas that affect their original thinking. Furthermore, it has been found that AI use affects original thinking, critical thinking, and creative thinking skills, but the direction and magnitude of this effect are related to individuals’ approaches to AI use and their individual characteristics.

The Adaptive Role in Developing Habits

The codes compiled in the theme examining the opinions on the transformation of AI use into a habit in difficult and academic tasks were discussed under two headings: (a) changes in intellectual and critical thinking depending on usage frequency and (b) adaptation of learning methods.

The findings under the heading of “changes in intellectual and critical thinking depending on usage frequency” indicate that AI use is frequently repeated in students.

The use of AI in academic tasks appears to have a positive impact on students’ perceptions of self-efficacy by supporting learning methods. Furthermore, students reported exhibiting more exploratory, deep, and innovative learning behaviors with AI.

The Transformative Role on Strategic Study Approaches

In this theme, which examines opinions on strategic approaches to using AI for challenging and academic tasks, the compiled codes are categorized under two headings: (a) validating self-awareness and self-regulation and (b) balancing diverse methods when working with artificial intelligence. Among the strategies adopted to manage cognitive processes and ensure productivity through conscious awareness and self-regulation when using AI technologies, codes such as critical thinking and verification/verification were obtained. In this context, students’ leading strategy in using AI is to first understand the subject themselves and then seek support from AI.

The codes examined under the heading of “balancing diverse methods” encompass two fundamental normative concerns regarding AI use: cognitive and ethical.

Practical Suggestions

The findings suggest that AI literacy in doctoral education should not be taught as a standalone technical skill. Instead, it should be integrated with metacognitive awareness, self-regulated learning, and cognitive effort management. Doctoral students need to learn not only how to use AI tools, but also how to monitor, question, verify, and justify their use during complex academic tasks. Students should be encouraged to compare AI responses with scholarly sources, evaluate their accuracy, explain their limitations, and reflect on how AI influenced their thinking. Such practices can support more critical and responsible AI use while reducing the risk of cognitive offloading and cognitive complacency.

Future studies should do:

Future research should examine the causal effects of AI use on perceived cognitive load, cognitive effort, and metacognitive regulation through experimental and mixed-methods designs. Longitudinal studies are also needed to understand how sustained AI use shapes learning habits, cognitive miserliness, and self-regulated learning over time. Also, future studies should include broader cognitive and ethical variables such as creative cognition, epistemic trust, ethical sensitivity, academic self-efficacy, learner autonomy, and intellectual ownership.