Education and Information Technologies

https://doi.org/10.1007/s10639-024-12966-8

The relationship between inert thinking and ChatGPT

dependence: An I-PACE model perspective

Jian-Hong Ye1,2 · Mengmeng Zhang3 · Weiguaju Nong4 · Li Wang5 ·

Xiantong Yang6

Received: 17 December 2023 / Accepted: 8 August 2024

© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature
2024

Abstract
ChatGPT, as an example of generative artificial intelligence, possesses high-level
conversational and problem-solving capabilities supported by powerful computa-
tional models and big data. However, the powerful performance of ChatGPT might
enhance learner dependency. Although it has not yet been confirmed, many teachers
and scholars are also concerned about this issue. Therefore, it is necessary to inves-
tigate this topic further. This study’s objective is to explore the association between
inert thinking, positive experiences with ChatGPT, avoidance learning motivation,
and ChatGPT dependence, based on the Interaction of Person-Affect-Cognition-
Execution (I-PACE) model. Employing a cross-sectional design, we conducted an
online survey with 870 Taiwanese university students, who had an average age of
22.81 years. The study found that inert thinking was positively associated with
both positive experiences with ChatGPT and ChatGPT dependence. Furthermore,
a significant association was found between inert thinking and avoidance learning
motivation. Positive experience with ChatGPT was also positively related to avoid-
ance learning motivation and ChatGPT dependence. Due to the scarcity of empiri-
cal research on generative artificial intelligence, the issues that people worry about
when discussing AI were confirmed in this study. Moreover, avoidance learning
motivation was positively correlated with ChatGPT dependence. Based on these
findings, this study calls for educators to help students overcome inert thinking and
avoidance learning motivation to prevent dependency on emerging technologies.

Keywords Artificial intelligence generated content (AIGC) · Avoidance learning

motivation · ChatGPT · Inert thinking · Interaction of the person-affect-cognition-
execution (I-PACE) model · Positive experience

Extended author information available on the last page of the article

13
 Education and Information Technologies

1 Introduction

ChatGPT is one of the most transformative AI tools developed in recent years, and
can significantly enhance productivity in various settings. However, such generative
AI tools also present numerous challenges in ethics, morality, and policy (Dwivedi et
al., 2023). Applications based on generative artificial intelligence have the potential
to revolutionize virtual learning (Ali et al., 2024). As most emerging technologies
are not yet widely adopted in classrooms, there is a need for more research that inte-
grates educational experience with user experience. This will enable educators to
understand how to use these technologies more effectively, avoid misuse, and make
teaching and learning more diverse and personalized (Ye et al., 2023b). From the
above, it can be seen that artificial intelligence-generated content (AIGC), including
ChatGPT, may have both positive and negative effects on education. ChatGPT is gen-
erative, meaning it can create new data, not just analyze existing data, especially in
the context of text (Dwivedi et al., 2023). In the educational context, ChatGPT helps
learners by retrieving, answering, hinting, planning, summarizing, embellishing, ana-
lyzing, and generating information, content, or questions. Moreover, it also offers
answer-oriented or planning support in everyday life. The inherent nature of AI and
chatbots is neutral; it is the potential negative impacts arising from their dependency
and use that warrant further attention and careful consideration (King & ChatGPT,
2023). Furthermore, the education sector has not yet fully understood the impact
of artificial intelligence tools such as ChatGPT and how they might integrate into
the learning process. Their impact on education should be carefully assessed so that
we can be vigilant about their potential effects on education (Elbanna & Armstrong,
2024). Based on this, the study was conducted to understand the situation of univer-
sity students’ ChatGPT dependence.
Within the theoretical framework of internet/media use, the Interaction of Person-
Affect-Cognition-Execution (I-PACE) model serves as an effective application frame-
work. Specifically, the I-PACE model provides a theoretical framework for internet
use disorders (Brand et al., 2016), where the “person” component refers to individual
characteristics. The subsequent two components represent affective and cognitive
responses to different situational triggers. Finally, these responses may produce final
executive functions, related to an individual’s inhibitory control and decision to use a
certain internet application (Hu et al., 2023). According to the I-PACE model, nega-
tive affective and cognitive responses to external or internal stimuli can promote
the use of the Internet and its maintenance (especially the craving component of the
I-PACE model) (Brand et al., 2016, 2019). Given the effectiveness and breadth of the
I-PACE model in explaining Internet media, it has quickly gained recognition and
application among scholars in cyberpsychology since its introduction. Prior studies
indicated that the I-PACE model is highly suitable for explaining maladaptive online
usage behaviors (Brand et al., 2019). Therefore, this study also employed the I-PACE
model as a theoretical framework to explain the formation process of dependency on
the use of ChatGPT.
While technology provides convenience in people’s lives, it can also potentially
make them less active. Tlili et al. (2023) suggested that one ethical challenge of using
ChatGPT might be the reduction in students’ critical thinking abilities. Additionally,

13
Education and Information Technologies

in discussion posts on social media, there are concerns that the widespread use of
AI tools may foster superficial learning habits and erode students’ social and critical
thinking skills (Mogavi et al., 2024). Although large language models can offer rela-
tively accurate and relevant information, they cannot replace the creativity, critical
thinking, or problem-solving skills developed through human guidance, as noted by
Dwivedi et al. (2023). Therefore, including cognitive thinking variables in ChatGPT
research is necessary. At the heart of inert thinking, a term for lazy thinking, is the
Dual Process Theory (Kahneman, 2011), which proposes that humans possess two
types of thinking styles: intuitive and analytical. Inert thinking, based on an intuitive
style, is seen as a barrier to reflection, deep thinking, and analysis, and is considered
detrimental to an individual’s growth and progress. Thus, although generative arti-
ficial intelligence has powerful auxiliary functions, relying on inert thinking may
contribute to adverse usage conditions. Therefore, this study took inert thinking as its
independent variable.
ChatGPT is a milestone in AI text generation, giving users a sense of their lives
being closer to AI for the first time (Ma & Huo, 2023). Its provision of specific and
relevant information about various subjects or topics makes it valuable to many users
in research (Tlili et al., 2023). ChatGPT ensures that users receive fast and timely
responses to their queries, thus improving their overall experience (Panda & Kaur,
2023). The findings of Kosov et al. (2023) emphasized that psychological responses
are precursors to the effectiveness of teaching and educational technologies, where
positive psychological responses increase effectiveness, while negative responses
decrease it. Therefore, when users receive positive feedback or appropriate responses
from their use of ChatGPT, it may create a positive experience. From the perspective
of the I-PACE model, a positive experience with ChatGPT will be an important fac-
tor; we therefore included this variable in the study.
Moreover, ChatGPT has the potential to completely transform the way we gener-
ate, distribute, and use instructional information in e-learning environments. Each
subject can be researched and learned by individual learners using ChatGPT (Javaid
et al., 2023). However, the powerful capabilities of ChatGPT also bring about many
instances of negative use. For example, college students may cheat on their paper
writing assignments by copying and pasting the responses generated by ChatGPT
into their own papers, providing answers to exam questions, generating fake data/
reports, or substituting the completion of reports (King & ChatGPT, 2023). These
actions not only raise concerns about academic integrity and plagiarism (Cotton et
al., in press), but also suggest that ChatGPT might lead researchers and students to
become lazy. The aforementioned underlying issues suggest that students are utilizing
ChatGPT not as a learning aid but rather as a tool to circumvent learning tasks. These
underlying issues indicate that students are not using ChatGPT as an aid to learn-
ing, but rather as a means to avoid learning tasks. Consequently, avoidance learning
motivation should be considered an important personal factor in the improper use of
ChatGPT, and thus was included as a variable in this study.
Furthermore, there is a worry that the powerful functions of AIGC are making
people increasingly reliant on it. If a given technology enhances an individual’s work
performance within an organization, choosing to utilize that technology is a rational
decision at both the individual and organizational levels. As a result, users may grav-

13
 Education and Information Technologies

itate towards advanced communication platforms, perceiving them as being more

user-friendly and beneficial compared to previous tools or media (Park et al., 2013).
Moreover, the more natural and human-like the responses of AI systems or chatbots,
the higher the probability that consumers will develop excessive dependence on them
(Baird & Maruping, 2021). This underscores that users of ChatGPT can success-
fully attain significant goals in their daily lives. If the features and characteristics of
ChatGPT contribute to its utility and ease of use, ultimately leading to widespread
adoption, individuals are likely to form a dependency on it. Therefore, this study
incorporated ChatGPT dependence as an outcome variable.
The majority of early studies primarily focused on engineering aspects, such as
developing new algorithms and enhancing machine learning/deep learning technolo-
gies (Chen et al., 2020). Therefore, Chen et al. have suggested that given the reinforce-
ment effect of artificial intelligence on education, we need more research to delve into
whether and how these emerging technologies can benefit education. Simultaneously,
this will enable educators to better understand how to utilize these technologies more
effectively and mitigate their potential misuse (Ye et al., 2023b). Overall, since Chat-
GPT is a new topic that only began to gain attention in 2023, while there are many
opinion articles, empirical research is still lacking. As a groundbreaking and iconic
generative artificial intelligence (GAI) product, studying it is meaningful for future
education and artificial intelligence research. Furthermore, teachers and researchers
in related fields are very concerned about the potential negative impacts of ChatGPT
on student users. They highlight problematic usage behaviors of artificial intelligence
in educational settings (Zhang et al., 2024).
To fill this research gap, this study aimed to explore the association among inert
thinking, positive experiences, avoidance learning motivation, and ChatGPT depen-
dence within the framework of the I-PACE model. Research findings not only enhance
comprehension of the broad applicability of the I-PACE model beyond addictive
entertainment media, but also provide a comprehensive explanation of factors related
to instrumental dependence. Moreover, these results help researchers and educators
gain deeper insights into the association between students’ positive interactions with
artificial intelligence and their reliance on AI technology, facilitating the develop-
ment of targeted strategies to mitigate the negative impacts of AI applications.

2 Theoretical models and hypotheses

2.1 I-PACE model

Brand et al. (2016) proposed that internet use disorders involve an interaction of
person-affect-cognition-execution. The I-PACE model aims to summarize processes
related to all types of specific internet use disorders (Young & Brand, 2017). Recently,
as a comprehensive approach, the I-PACE model originated from and integrates cur-
rent theories essential to explaining substance use disorders and behavioral addictions
(Brandtner et al., in press). This model encompasses several key elements. First, it
encompasses personal characteristics such as personality traits, social cognition, and
cognitive vulnerability. The second element includes emotional factors influencing

13
Education and Information Technologies

individual behavior, encompassing emotional responses to various stimuli and cor-

responding coping strategies. The third element involves shaping personal cognitive
processes and biases. The final element pertains to actual behaviors, such as decision-
making in Internet applications (Zhang et al., 2024). Dependency habits emerged as
a result of the interplay between these core elements. In sum, a central idea of the
I-PACE model is that the development of problematic and addictive behaviors only
occurs in the interaction between an individual’s susceptibility variables and certain
aspects offered by specific situations. These interactions lead to satisfying and com-
pensatory experiences related to particular behaviors (Brand et al., 2019).
From this, it is evident that under the framework of the I-PACE model, through
key elements such as a person, affect, cognition, and execution, one can effectively
explain people’s improper use of media. Meanwhile, instrumental and recreational
uses share similar psychological mechanisms, so the I-PACE model can be used
not only to explain the overuse/misuse/addiction of recreational media, but also the
overuse of instrumental media. Furthermore, according to the characteristics of the
model, the I-PACE model is useful for understanding the determinants of artificial
intelligence dependence. As ChatGPT is an instrumental medium, this study used
the I-PACE model as its theoretical framework to explain the association between
college students’ ChatGPT usage experiences and the association between learning
motivation and instrumental use dependence.

2.2 Research hypotheses

Intuitive thinking is often described as part of the reflexive system, characterized as

“intuition” and “experience” or “pattern recognition,” triggering automatic modes of
thought (Tay et al., 2016). This reflexive, open cognitive style denotes a “lazy” deci-
sion-making approach, where a wide range of propositions are accepted uncritically,
regardless of their epistemic value (Pennycook & Rand, 2020). From the perspective
of the I-PACE model, cognitive responses lead to decision-making in a particular
way (Brand et al., 2019). Users utilizing ChatGPT are expected to produce more
accurate and reliable results, as ChatGPT can provide precise answers to various
questions and topics (Hariguna & Ruangkanjanases, 2023). Therefore, inert thinking
might influence people’s experiential perceptions of using a tool.
Convenience is a situational criterion for people’s choices and actions in the pro-
cess of seeking information (Connaway et al., 2011), and the development of technol-
ogy encourages lazy behaviors or addiction to convenience (Singharat et al., 2023).
The Lazy User Theory (LUT) emphasizes that users generally choose solutions that
satisfy their (information) needs with the least effort (lazy user behavior) (Pinto et
al., 2019). From the perspective of the I-PACE model, the perception of internal
or external triggers may promote the experience of prompt response and desire or
craving, as a reaction to these cues (Brandtner et al., in press). Therefore, users with
inert thinking, by solving tasks through efficient ChatGPT, are more likely to have a
positive experience. This shows that cognition (thinking) is an important influencing
variable (Elhai et al., 2020). Integrating the perspectives of the I-PACE model and
LUT, when users use ChatGPT with an inert thinking pattern, they are likely to have
a better experience and a stronger dependence on ChatGPT; moreover, users with a

13
 Education and Information Technologies

higher level of inert thinking may also have a higher avoidance learning motivation
for learning tasks. Therefore, this study explored the association between inert think-
ing and participants in terms of positive experience, avoidance learning motivation,
and ChatGPT dependence, with the following hypotheses:

H1: Inert thinking is positively correlated with positive experience.

H2: Inert thinking is positively correlated with avoidance learning motivation.

H3: Inert thinking is positively correlated with ChatGPT dependence.

According to the I-PACE model, emotional stimuli may promote internet use (Brand
et al., 2016, 2019). Internet use can bring about a kind of positive experience, which
strengthens our attention bias and craving for internet use. This experience also solid-
ifies cognitive biases and coping methods related to the internet, and these consolida-
tion mechanisms lead us to repeatedly engage in behaviors such as social media use
(Huang et al., 2020). Therefore, the positive experience and satisfaction gained from
using the internet drive users to continue its use (Wang et al., 2023). As a tool that can
enhance efficiency and productivity, ChatGPT can help students enrich their knowl-
edge in challenging tasks and controversial topics, bringing about positive emotional
experiences. However, excessive use may lead to researchers’ and students’ depen-
dency behaviors (Qasem, 2023). Hence, we infer that the positive emotional stimuli
from using ChatGPT may also prompt users to develop a dependency on it.
Moreover, individuals’ experiences are related to the approach or avoidance learn-
ing motivation. People continuously strive to obtain pleasant stimuli and do their
utmost to avoid painful stimuli (Howard & Smith, 2023). Avoidance learning moti-
vation can be defined as the motivation or direction of behavior driven by negative
stimuli (objects, events, possibilities) (Elliot, 2006). Therefore, emotions not only
change the tendency to approach or avoid targets, but also affect responses to stimuli
related to those targets (Howard & Smith, 2023). In other words, the emotional expe-
rience derived from using ChatGPT may lead students to develop a motivational ten-
dency to avoid tasks. It can be inferred that if users gain a positive experience from
using ChatGPT, it might increase their motivation to avoid learning tasks and their
dependence on ChatGPT. Thus, this study explored the relationship between positive
experience and participants in terms of avoidance learning motivation and ChatGPT
dependence, with the following hypotheses:

H4: Positive experience is positively correlated with avoidance learning motivation.

H5: Positive experience is positively correlated with ChatGPT dependence.

H6: Avoidance learning motivation is positively correlated with ChatGPT

dependence.

All organisms have an innate response to approach potentially beneficial stimuli

and avoid potentially harmful ones (Su et al., 2022). Students with a high avoidance

13
Education and Information Technologies

learning motivation actively avoid negative behaviors in learning (Hong et al., 2019;
Ye et al., 2023b). Avoidance learning motivation is considered a stress response, and
this pattern of response can lead to addiction (Thomas et al., 2011). This also explains
why the misuse of ChatGPT as a tool for plagiarism is believed to lead researchers to
become dependent on this technology (Qasem, 2023). It is evident that the more users
want to avoid their learning tasks, the higher their dependence on ChatGPT might be.
Therefore, this study explored the associations between avoidance learning motiva-
tion and ChatGPT dependence, with the following hypothesis:

H6: Avoidance learning motivation is positively correlated with ChatGPT

dependence.

2.3 Hypothesis model

An efficient model framework needs to ensure simplicity and universality, charac-

teristics that the I-PACE model embodies. The I-PACE model consists of a process
model (Young & Brand, 2017) and is a general approach to the potential processes of
addictive behavior, and hence does not specify the role and position of each particular
cognitive or affective process (Brandtner et al., in press). The I-PACE model com-
prehensively analyzes inappropriate usage behavior through the interaction between
vulnerability variables, emotional and cognitive responses to specific stimuli, and
executive functions (Sun et al., 2024). That is to say, in the I-PACE model, personal
perceptions influence people’s states and cognitive reflections, ultimately changing
their thoughts about actions. Therefore, the variables used in the I-PACE model vary
for different research topics (Nong et al., 2023). Overall, this model is a theoretical
framework that describes the processes involved in the development and dependent
use of specific internet applications. Therefore, this research mentioned above indi-
cates that the I-PACE model is highly suitable for studying ChatGPT dependence.
Thus, in this study, the I-PACE model was chosen to explain the interactions of
cognitive (inert thinking), affective (positive experience with ChatGPT), personal
(avoidance learning motivation), and executive behavior (ChatGPT dependence) fac-
tors in ChatGPT use addiction. The combination of these four variables seems to
align with the mechanisms identified in the I-PACE model. The hypothesis model is
illustrated in Fig. 1.

3 Research design

3.1 Procedure

This study adopted an online questionnaire survey based on a cross-sectional design.

Convenience sampling was used, and Google Forms was utilized to distribute ques-
tionnaires to university students in Taiwan. The data collection period was from July
28 to September 12, 2023, with a total of 1,062 questionnaires collected. An informed
consent statement was included in the electronic questionnaire, and participants were

13
 Education and Information Technologies

Fig. 1 Hypothesis model

Table 1 Demographic Variable Number

information
Gender Male: 554 participants (63.7%)
Female: 316 participants (36.3%)
Education System Undergraduate students: 505 (58%)
Master’s students: 312 (35.9%)
Ph.D. students: 53 (6.1%)
School Affiliation Public schools: 689 participants (79.2%)
Private schools: 181 participants (20.8%)
Type of School Universities of Science and Technology: 504
participants (57.9%)
General Universities: 366 participants (42.1%)
Usage Experience Less than 1 month: 29 participants (3.3%)
1 to 3 months: 250 participants (28.7%)
3 to 6 months: 428 participants (49.2%)
More than 6 months: 163 participants (18.7%)

informed about the study and their role in it before filling out the questionnaire. Thus,
electronic informed consent was obtained from all participants involved in the study.

3.2 Participants

The number of participants in this study was 1,062. Based on lie detector questions,
response times less than 3 min, and incomplete responses, 192 questionnaires were
deemed invalid, leaving 870 valid participants with an effective recovery rate of
81.9%. Basic information about the participants, including gender, education system,
school affiliation, type of school, and experience using ChatGPT, was collected, with
specific results shown in Table 1. Additionally, the average age of participants was
22.81 years, with a standard deviation of 3.36 years.

13
Education and Information Technologies

3.3 Measures

All questionnaires in this study were self-reported, using a Likert 5-point scale for
assessment, where 1 to 5 represent degrees of disagreement to agreement. There was
a total of 37 questions in the questionnaire, including six background variable ques-
tions, one lie detection question, and 30 questions related to the four model variables.
Before distribution, the questionnaire was reviewed by three experts in educational
technology to ensure its expert validity.

3.3.1 Inert thinking

The inert thinking scale, comprising eight questions, was developed based on the
concept of inert thinking to measure participants’ perceptions of their reluctance to
think deeply during thinking or task execution. Example questions include: “I do not
like to delve deeply into the causes of problems.” A higher average score on this vari-
able indicates a tendency towards laziness, distraction, and reluctance to think deeply
during thought processes.

3.3.2 Positive experience

The positive experience scale, comprising seven questions, was adapted and revised
from Hwang et al.’s (2019) value scale. Hwang et al. confirmed users’ learning effec-
tiveness assistance experience with educational games. In this study, to better align
with the research theme, this definition is termed as a positive experience. It mea-
sures the extent to which participants perceive the positive effects of using ChatGPT.
Example questions include: “The content answered by ChatGPT can be applied to
my academic learning,” and “ChatGPT helps me learn knowledge more quickly.”
A higher average score on this variable indicates a higher degree of positive experi-
ences felt by the participant since using ChatGPT.

3.3.3 Avoidance learning motivation

The avoidance learning motivation scale, consisting of eight questions, was adapted
and revised by Ye et al. (2023b). It measures the level of motivation participants have
to avoid learning tasks. Example questions include: “I often think about avoiding
learning tasks,” and “When I do homework, I choose the easiest method possible.” A
higher average score on this variable indicates a greater tendency of the participant to
avoid or take a perfunctory approach to learning tasks.

3.3.4 ChatGPT dependence

The ChatGPT dependence scale, comprising seven questions, was developed based
on the concept of instrumental use dependence. It measures the level of participants’
dependence on using ChatGPT. An example question is: “I find it difficult to stop
using ChatGPT.” A higher average score on this variable indicates a higher level of
instrumental dependence on using ChatGPT.

13
 Education and Information Technologies

3.4 Statistical tools

Structural Equation Modeling (SEM) is a multivariate data analysis method employed

to examine intricate relationships between latent constructs and observed indicators
(Hair et al., 2021). It is a widely used method to test substantive hypotheses in psy-
chology and other social sciences (Jobst et al., 2023). SEM has various develop-
ments that make it increasingly useful in a variety of data types, research designs,
research questions, and research contexts, such as one of SEM’s key purposes being
the evaluation of theoretical relationships (Zyphur et al., 2023). As a powerful tool
for hypothesizing theoretical causal relationships, SEM is highly regarded by many
quantitative researchers. Therefore, this study used the AMOS 26.0 software based
on SEM technology to validate the research hypothesis model.

4 Results

4.1 Common method bias (CMB)

Due to the potentially severe impact of common method biases on research outcomes,
understanding their sources and when they are particularly likely to become an issue
is crucial. One of the most widely used techniques by researchers to address the
common method variance problem is Harman’s single-factor test (Podsakoff et al.,
2003). Therefore, this study conducted an analysis using Harman’s single-factor test,
where the first factor accounted for 40.32% of the variance, lower than the threshold
suggested by Korsgaard and Roberson (1995) and Podsakoff et al. (2003), who stated
that if an Exploratory Factor Analysis (EFA) involving all the main research variables
results in a single factor accounting for more than 50% of the variance, a common
method bias (CMB) issue exists. This implies that the data in this study did not have
a common method bias problem.

4.2 Measurement model

Before conducting the validation of the research model, an analysis of the fit of each
measurement model was necessary to ensure the congruence of the variable content.
During this stage, the root mean square error of approximation (RMSEA), goodness
of fit index (GFI), adjusted goodness of fit index (AGFI), and factor loadings (FL)
of items were examined. Following the recommendations by Hair et al. (2019), a
measurement model analysis and FL analysis were conducted, as well as the elimina-
tion of items (as per the standards outlined in Table 2). The results showed that the
goodness-of-fit indices in this study aligned with the recommendations by Hair et
al. (2019). In terms of external validity analysis, the t-values met the standards sug-
gested by Cor (2016) and Green and Salkind (2004). Furthermore, during the analysis
of the measurement model, the Inert Thinking scale was reduced from eight items
to five, the Positive Experience scale from seven items to five, and the ChatGPT
Dependence scale from seven items to six. During this stage, items 1, 2, and 5 of Inert
Thinking were removed, as well as items 5 and 6 of Positive Experience, and item

13
Education and Information Technologies

Table 2 Measurement model results

Fit Indices Threshold Inert thinking Positive Avoidance learn- ChatGPT
Values experience ing motivation dependence
χ2 --- 13.4 15.6 41.4 43.6
df --- 5 5 20 9
χ2/df <5 2.68 3.12 2.07 4.84
RMSEA < 0.10 0.04 0.05 0.04 0.07
GFI > 0.80 0.99 0.99 0.99 0.98
AGFI > 0.80 0.98 0.98 0.98 0.96
FL > 0.50 0.58 ~ 0.79 0.69 ~ 0.73 0.57 ~ 0.80 0.69 ~ 0.77
t >3 25.18 ~ 41.06 23.34 ~ 25.24 22.13 ~ 43.03 25.1 ~ 29.82

Table 3 Reliability and validity

Construct M SD α CR AVE FL
--- --- > 0.70 > 0.70 > 0.50 > 0.50
Inert thinking 3.36 0.93 0.86 0.86 0.55 0.74
Positive experience 3.77 0.71 0.84 0.84 0.51 0.71
Avoidance learning motivation 3.32 0.89 0.91 0.91 0.55 0.74
ChatGPT dependence 3.81 0.65 0.86 0.86 0.51 0.71

4 of ChatGPT Dependence. Additionally, it is worth noting that avoidance learning

motivation was not subject to item reduction in this analysis, as shown in Table 2.

4.3 Reliability and validity

For a quantitative research study to yield reliable and trustworthy results, it needs to
demonstrate acceptable levels of reliability and validity. As shown in Table 3, in this
study, the Cronbach’s α values and Composite Reliability (CR) values for each vari-
able ranged from 0.84 to 0.91, while Factor Loading (FL) values ranged from 0.71
to 0.74, and the Average Variance Extracted (AVE) values ranged from 0.51 to 0.55.
These results aligned with the standards proposed by Hair et al. (2019), indicating
acceptable numerical values. Additionally, the variables had mean scores ranging
from 3.32 to 3.81, with standard deviations ranging from 0.65 to 0.93.
Construct Discriminant Validity analysis can assist researchers in identifying the
absence of multicollinearity issues among variables. If the correlation coefficient is
smaller than the square root of the Average Variance Extracted (AVE), then the dis-
criminant validity of each variable is considered acceptable, as outlined by Cheung
et al. (in press). Additionally, regarding the discriminant validity analysis, Ahmad
et al. (2016) also proposed that another assessment criterion is that the correlations
between constructs should be less than 0.85. As seen in Table 4, the analysis results
of this study met the recommendations of both scholars.

4.4 Overall structural model fit results

To ensure the adequacy of the research model, after conducting measurement model
analysis and reliability and validity assessments, an overall structural model fit evalu-

13
 Education and Information Technologies

Table 4 Construct discriminant Construct 1 2 3 4

validity
1. Inert thinking (0.74)
2. Positive experience 0.39 (0.71)
3. Avoidance learning motivation 0.73 0.44 (0.74)
4. ChatGPT dependence 0.52 0.53 0.55 (0.71)

Fig. 2 The result model of ChatGPT dependence

ation is necessary for model validation. During this stage, eight indicators, RMSEA,
GFI, AGFI, normed fit index (NFI), non-normed fit index (NNFI), comparative fit
index (CFI), incremental fit index (IFI), and relative fit index (RFI), should all be
higher than 0.80. Additionally, two indicators, parsimonious normed fit index (PNFI)
and parsimonious goodness of fit index (PGFI), should be higher than 0.50. The fit
indices for this study were as follows: χ² = 425.6, df = 246, χ²/df = 1.73, RMSEA = 0.03,
GFI = 0.96, AGFI = 0.95, NFI = 0.96, NNFI = 0.98, CFI = 0.98, IFI = 0.98, RFI = 0.96,
PNFI = 0.86, and PGFI = 0.79. These results aligned with the acceptable numerical
standards proposed by Abedi et al. (2015) and Hair et al. (2019).

4.5 Structural model validation results

The validation of this study, conducted using the AMOS software, revealed the fol-
lowing relationships: Inert thinking had a positive correlation with positive experi-
ence (β = 0.45***); inert thinking had a positive correlation with avoidance learning
motivation (β = 0.75***); inert thinking had a positive correlation with ChatGPT
dependence (β = 0.20**); positive experience had a positive correlation with avoid-
ance learning motivation (β = 0.16***); positive experience had a positive correlation
with ChatGPT dependence (β = 0.41***); and avoidance learning motivation had a
positive correlation with ChatGPT dependence (β = 0.25***). These validation results
supported all seven research hypotheses, as illustrated in Fig. 2.

13
Education and Information Technologies

Furthermore, the explanatory power of inert thinking for positive experience was
20%, with an effect size (f²) of 0.25; the explanatory power of inert thinking and posi-
tive experience for avoidance learning motivation was 69%, with an effect size (f²)
of 2.23; and the explanatory power of inert thinking, positive experience, and avoid-
ance learning motivation for ChatGPT dependence was 53%, with an effect size (f²)
of 1.13, as shown in Fig. 2.

4.6 Indirect effects analysis

Through bootstrap analysis, this study found that inert thinking positively but indi-
rectly predicted both avoidance learning motivation and dependence on ChatGPT.
Additionally, positive experiences had a positive indirect effect on ChatGPT depen-
dence, as shown in Table 5.

4.7 Discussion

4.7.1 Main findings

In the I-PACE model, perceptions of internal and external triggering factors are
specified to describe how stimuli are related to individuals’ emotional and cognitive
responses. Individuals may perceive external or internal triggering factors in specific
situations, and these perceptions may result in emotional and cognitive responses,
such as increased attention to these stimuli and impulses to act in specific ways.
In this study, within the framework of the I-PACE model, a hypothesis model was
validated, and the theoretical associations of six hypothesis paths were confirmed. In
other words, from the perspective of the I-PACE model, the cognitive factors (inert
thinking) of student users are associated with emotional factors (positive experiences
of using ChatGPT), as well as personal factors (avoidance learning motivation) and
behavioral execution (dependence on ChatGPT) factors, while emotional factors are
also linked to personal and behavioral execution factors.
In summary, the main findings of this study are as follows: (1) The I-PACE model
can be used to explain dependence on ChatGPT; (2) Inert thinking negatively affects
individuals’ motivation for learning; (3) Improper use of ChatGPT may increase
dependency; (4) Avoidance learning motivation is significantly associated with
dependence on ChatGPT; and (5) There are indirect associations between inert think-
ing, avoidance learning motivation, and ChatGPT dependence. Likewise, positive
experience also shows indirect associations with ChatGPT dependence.

Table 5 Indirect effects analysis Construct Inert thinking Positive

experience
β 95% CI β 95%
CI
Avoidance learning 0.07*** [0.04,
motivation 0.10]
ChatGPT dependence 0.39*** [0.27, 0.04*** [0.02,
0.52] 0.07]

13
 Education and Information Technologies

4.7.2 Inert thinking is associated with positive experience, avoidance learning

motivation, and ChatGPT dependence

This study posited that inert thinking might affect individuals’ experiential percep-
tions of using tools, aligning with the theoretical concept proposed by Pinto et al.
(2019). The study found that inert thinking has a positive correlation with positive
experience (H1), avoidance learning motivation (H2), and ChatGPT dependence
(H3). This is consistent with the perspective that users generally choose solutions
that fulfill their (information) needs with minimal effort (lazy user behavior), as sug-
gested by the Lazy User Theory (LUT). Additionally, since ChatGPT can provide
precise answers to a wide range of questions and topics, users are expected to achieve
more accurate and reliable results (Hariguna & Ruangkanjanases, 2023). According
to the I-PACE model, cognitive responses influence decision-making in specific ways
(Brand et al., 2019). Therefore, users with inert thinking are more likely to have posi-
tive experiences when using ChatGPT efficiently to accomplish tasks, which may, in
turn, be positively related to ChatGPT dependence. Thus, it can be observed that inert
thinking is a crucial antecedent variable affecting users of AIGC. In addition, from
the results of the study, we can see that artificial intelligence is a double-edged sword;
the application of artificial intelligence has many benefits, but it is also accompanied
by potential risks, depending on the user’s purpose and motivation to apply it.
Additionally, conversations with ChatGPT provided insights into these validation
results. It suggested that the positive relationship between inert thinking and positive
experience could be explained by factors such as “ease of problem-solving,” “quick
feedback,” “confidence and a sense of success,” and “convenience and real-time sat-
isfaction.” However, it also cautioned that ChatGPT should be used with caution
to ensure it does not hinder individual active learning and problem-solving abili-
ties. The positive association between inert thinking and ChatGPT dependence can
be explained by factors such as “convenience and real-time satisfaction,” “lack of
active thinking,” “technological dependence,” and “psychological stress.” In sum-
mary, users with inert thinking are more likely to have positive experiences when
using ChatGPT due to its powerful functionality and human-like interaction, leading
to a higher likelihood of ChatGPT dependence. Furthermore, individuals with inert
thinking may be less engaged in learning and may seek quick and simple completion
of learning tasks, or even avoid them altogether. These results align with the theoreti-
cal assumptions of the I-PACE model.

4.7.3 Positive experience with ChatGPT is positively related to avoidance learning

motivation and ChatGPT dependence

Despite the advantages that AI brings to education, its adoption has raised practical
issues. Among these concerns, the decline in student motivation and the over-reli-
ance on AI tools are particularly prominent (Morales-García et al., 2024). This study
found a positive correlation between Positive Experience with ChatGPT and avoid-
ance learning motivation (H4), which aligns with the perspective of Howard and
Smith (2023). They argued that individuals continuously strive to seek pleasurable
stimuli and avoid painful stimuli. A positive tool usage experience may help students

13
Education and Information Technologies

understand effective ways to avoid learning tasks. Many educators have also found
that ChatGPT’s large language models can assist university students in various lan-
guage and writing tasks across subjects, helping them better understand and analyze
materials. This observation is in line with the perspective presented by Judijanto et al.
(2024), that is, artificial intelligence technology has significant potential for enhanc-
ing learning efficiency and delivering improved personalized learning experiences
to students. Therefore, when users have positive experiences with ChatGPT, it not
only assists them in their learning, but may also be correlated with a higher degree
of avoidance learning motivation. While ChatGPT offers a range of advantages, it
may inadvertently contribute to excessive dependency (Van Dis et al., 2023), thereby
impeding students’ perceptions of their independent writing abilities (Bouzar et al.,
2024). Specifically, when student users experience the powerful effects of ChatGPT
during use, they may become more inclined to avoid completing learning tasks using
ChatGPT. There are certainly studies that show that AIGC can help learners develop
reflective and critical thinking, creativity, problem-solving skills, and conceptual
understanding (Rodrigues Vasconcelos & P dos Santos, 2023). Therefore, the out-
comes that result from the use of experiences are important.
Moreover, during conversations with ChatGPT, it provided insights into these ana-
lytical results. It suggested that the positive relationship between avoidance learning
motivation could be explained by factors such as “consistency of learning goals,”
“diversity of learning motivations,” “perceived learning outcomes,” and “emotional
support.” In other words, when users have positive experiences with ChatGPT while
assisting in the learning process, they are more likely to be motivated to avoid learn-
ing tasks using this tool. Additionally, this study found that Positive Experience
with ChatGPT also had a positive correlation with ChatGPT Dependence (H5). This
aligns with the I-PACE framework proposed by Brand et al. (2016) and Brand et al.
(2019), suggesting that emotions can promote internet use. Positive experiences with
technology can enhance attentional bias and craving for technology use, leading to
cognitive biases and coping mechanisms related to technology. These reinforcement
mechanisms are associated with repetitive engagement in technology use behaviors
(Huang et al., 2020). Thus, if users perceive ChatGPT as a tool that enhances effi-
ciency and productivity, they are more likely to adopt it (Ma & Huo, 2023). However,
this also comes with the risk of ChatGPT dependence (Qasem, 2023). Therefore, stu-
dents’ experiences with the use of AIGC tools not only make them more receptive of
this tool, but may also be related to technological dependence. Additionally, because
of the effective experiences with the technology, they may develop tendencies to
avoid learning tasks.
Furthermore, during conversations with ChatGPT, it provided insights into these
results and acknowledged the positive relationship of Positive Experience with Chat-
GPT and ChatGPT Dependence. This could be explained from perspectives such as
“technological dependence and satisfaction,” “success experiences,” “time and effort
savings,” and “convenience.” In other words, when users have a positive usage expe-
rience with ChatGPT while using it as an assistive learning tool, their dependence on
it is likely to increase. Therefore, these results align with the theoretical viewpoint of
the I-PACE model.

13
 Education and Information Technologies

4.7.4 Avoidance learning motivation is positively related to ChatGPT dependence

In this study, it was found that avoidance learning motivation is positively correlated
with ChatGPT dependence (H6). This aligns with the perspective of Thomas et al.
(2011), who argued that avoidance learning motivation, as a response to stress, is
part of an inappropriate avoidance coping pattern, which can be related to addic-
tion. Additionally, Qasem (2023) suggested that the misuse of ChatGPT as a tool for
plagiarism could increase individuals’ technological dependence. In other words, if
students develop thoughts of using AIGC tools to evade learning tasks, there is a high
likelihood of developing psychological dependence on the technology. These find-
ings also align with the perspectives of Hong et al. (2019); Ye et al. (2023b), indicat-
ing that students with high motivation to avoid learning will actively steer clear of
negative behaviors during the learning process. In addition, Simons et al. (2004) sug-
gested that more adaptive goal orientation will generate higher intrinsic motivation,
leading to the use of more adaptive cognitive strategies and better learning habits.
Moreover, during our conversations with ChatGPT, it provided insights into
these results and acknowledged the positive relationship between avoidance learn-
ing motivation and ChatGPT dependence. This relationship may be explained from
perspectives such as the “convenience of ChatGPT,” users’ “anxiety or stress,”
“low self-confidence,” “instant gratification, and unwillingness to attempt.” In other
words, when students have a higher level of motivation to avoid learning tasks, their
dependence on ChatGPT as a tool is likely to be higher. These findings are in line
with the propositions of the I-PACE model.

4.7.5 The intrinsic associations between variables

This study found that inert thinking has a positive association with avoidance learn-
ing motivation through positive experiences. Wirth et al. (2022) suggested that inert
thinking is a personality trait, leading individuals to invest less effort in their work.
The lazy user theory (Pinto et al., 2019) indicates that human behavior tends to fol-
low the path of least cognitive effort. According to ChatGPT’s opinion, when indi-
viduals adopt an inert mindset, they may avoid challenging activities or tasks due to a
lack of motivation to solve problems. However, occasional positive experiences can
reinforce this tendency by showing that positive learning outcomes are achievable
with the help of these tools, further encouraging the avoidance of challenges.
Moreover, this study revealed that inert thinking indirectly correlated with Chat-
GPT dependence through positive experiences and avoidance learning motivation.
The Lazy User Theory suggests that people make decisions by choosing the solution
that requires the least amount of effort to achieve their goals (Pinto et al., 2019).
This study’s findings are consistent with Ahmad et al. (2023), who noted the increas-
ing reliance of teachers and students on AI applications for tasks and assignments.
According to ChatGPT, inert thinking may incline individuals to avoid situations
with potential negative consequences. However, as they realize the positive outcomes
and experiences facilitated by tools like ChatGPT, they may reduce their reliance on
avoidance learning motivation. In essence, inert thinking amplifies positive experi-

13
Education and Information Technologies

ences and diminishes avoidance learning motivation, indirectly strengthening indi-

viduals’ motivation to depend on ChatGPT.
This study also found that positive experiences have a positive correlation with
ChatGPT dependence through avoidance learning motivation. This finding aligns
with Ahmad et al. (2023), who argued that due to the significant advantages of arti-
ficial intelligence, teachers and learners may lose interest in engaging in activities or
learning new skills or methods, thus becoming more reliant on AI systems. According
to ChatGPT, individuals who experience positive outcomes while using the platform,
such as successfully solving problems or obtaining useful information, are likely to
have reduced motivation to avoid learning. They begin to recognize that using these
tools can increase their positive outcomes, not just the negative consequences. As
avoidance learning motivation decreases, individuals may become more proactive in
relying on ChatGPT to address challenges or fulfill their needs, thus strengthening
their dependence on the platform.

4.7.6 Research limitations and future research directions

This study has some limitations that need to be addressed. First, it employed a cross-
sectional design, which permitted only the determination of correlations between
variables, rather than causal relationships. Consequently, future research should uti-
lize longitudinal or experimental designs to establish causal relationships between
variables. Meanwhile, this research was a self-report-based confirmatory study.
Although lie detection questions were included in the questionnaire, future studies
could incorporate objective measures to analyze the associations between subjective
and objective variables. Second, this study’s framework was built upon the I-PACE
model; however, some simplification occurred during the construction of the research
model, omitting the interactive aspects. The human-like characteristics of AI are con-
sidered capable of providing users with a high level of human-computer interaction
experience, which is also the most distinguishing feature of AI technology compared
to other technologies. Therefore, future work can enhance the AI dependency model
by incorporating the interactive features defined in the I-PACE model. Third, consid-
ering theoretical support, it may be worthwhile to address issues of outcome biases
caused by societal expectations or user characteristics by incorporating appropriate
control variables into the model. Fourth, expanding the I-PACE model to provide a
more detailed explanation of media usage dependence could enhance the persuasive-
ness of the research findings.
Furthermore, this study aimed to explore the relationship between ChatGPT
dependence and avoidance learning motivation. Future research should investigate
the associations between different usage motivations and experiences (positive, nega-
tive, or neutral), including approach and avoidance learning motivation, and AIGC
dependence to further enhance our understanding of the topic.
Additionally, assessing the usage experience of generative artificial intelligence
is crucial. While this study included several measurement questionnaires related to
usage experience, there is still room for improvement in systematically evaluating
the usage experience of artificial intelligence tools. Future research should be encour-
aged to develop reliable and valid assessment scales for the usage experience of tools

13
 Education and Information Technologies

such as AIGC. Additionally, there is a need to develop theoretical models or guiding

frameworks specifically tailored to AIGC-related phenomena. This will contribute to
advancing the research on technology applications.
Finally, AIGC has raised new educational considerations, but its widespread use is
relatively recent. Therefore, scholars need to continue exploring its effectiveness and
potential negative impacts through empirical studies. Additionally, as new technolo-
gies emerge, it is crucial to understand their contributions and challenges within edu-
cational systems to harness technology for educational empowerment. This involves
further exploration of AIGC’s role in supporting scientific research, enhancing tech-
nological writing effectiveness, and its various applications in education for stake-
holders such as students, teachers, and researchers.

5 Conclusion and contributions

5.1 Conclusion

The purpose of this study was to explore the associations between inert thinking,
positive experience with ChatGPT, avoidance learning motivation, and ChatGPT
dependence within the framework of the I-PACE model. The study findings are as
follows: (1) Inert thinking is positively associated with positive experiences with
ChatGPT and ChatGPT dependence. Inert thinking also has a close relationship with
avoidance learning motivation; (2) Positive experience with ChatGPT is positively
correlated with avoidance learning motivation and ChatGPT dependence; (3) Avoid-
ance learning motivation is positively correlated with ChatGPT dependence; and (4)
Inert thinking is associated with ChatGPT dependence through positive experience
and avoidance learning motivation. Inert thinking is also linked with avoidance moti-
vation through positive experience.
In this study, we also attempted to seek support from ChatGPT in interpreting
the results. However, it was observed that not all responses provided by ChatGPT
were highly accurate, and some answers were ambiguous. We needed to repeatedly
adjust the phrasing of our conversation sentences to obtain reasonably valid answers.
Importantly, while ChatGPT can assist in technological writing, users should evalu-
ate the correctness and effectiveness of its responses to avoid being misled.
Additionally, the powerful capabilities of generative artificial intelligence have
raised concerns, and this study confirms the importance of these concerns. As AI
and chatbots continue to advance, we must carefully consider their impacts and take
measures to prevent their misuse. When individuals use ChatGPT in an inert thinking
manner, they may become less inclined to enjoy learning and may develop excessive
dependence on AI tools. Based on the research findings, this study emphasizes the
importance of harnessing tools rather than relying on them. Overreliance on tools
does not promote information literacy; in fact, it can be detrimental to users’ informa-
tion literacy.

13
Education and Information Technologies

5.2 Research contributions

This study has made several contributions to the field. First, it provides evidence for
the application of the I-PACE model framework to explain tool-based usage depen-
dence, not just entertainment-based addiction behavior. This supports future research
in using this model to effectively explain tool-based usage dependence or overuse.
Meanwhile, the study also contributes as a reference basis for scholars interested in
the issue of instrumental dependence, regarding the application of the model in dif-
ferent fields and topics.
Second, the study confirms a significant correlation between inert thinking and
students’ positive experiences with ChatGPT. This finding underscores the impor-
tance of viewing technology as a tool rather than a dependency: tools should be used
to save time and effort, not to replace critical thinking. This contribution will help
educators better understand how to cultivate students’ abilities and thinking in the use
of digital tools. Preventing inert thinking among users is crucial, and teachers should
guide students to use AIGC tools, including ChatGPT, as part of formal and informal
learning, while avoiding direct reliance on them for generating answers or writing
reports, as this may not be an effective learning method. In other words, cultivating
students’ digital literacy, information literacy, and academic ethics awareness is cru-
cial in the era of digital transformation.
Lastly, this study highlights the significant potential of ChatGPT in academic
research, offering unprecedented assistance in data or text analysis, and summarizing
the key points of the literature. The study also emphasizes the ethical considerations
when using ChatGPT in research, as well as the importance of verifying the accuracy
and ethical standards of AI tool responses. It also calls for further exploration within
the boundaries of use on how to effectively utilize AIGC to assist academic research,
thereby achieving more efficient and higher-quality completion of academic research
and papers/reports. This helps researchers understand how to efficiently use AIGC
to present argumentative content in papers, thus shortening the overall research
duration.
Overall, this research provides insights into the relationships between thinking
styles, experiences with AI, motivation, and dependence, and offers recommenda-
tions for responsible AI usage and education.

Appendix 1 Validity analysis of items

No. Items M SD FL
Inert thinking
1 I like to do things based on my past personal experiences. 3.59 1.080 0.60
2 I like to refer to others’ experiences when doing things. 3.33 1.136 0.77
3 I do not like observing new things around me. 3.27 1.217 0.79
4 I do not like to delve deeply into the causes of problems. 3.30 1.216 0.77
5 I do not like to generalize or summarize problems. 3.32 1.175 0.77
Positive experience

13
 Education and Information Technologies

No. Items M SD FL
1 ChatGPT makes it easier for me to learn knowledge. 3.81 0.911 0.71
2 ChatGPT helps me learn knowledge more quickly. 3.72 0.863 0.73
3 ChatGPT helps me learn the relevant knowledge I need. 3.80 0.900 0.73
4 The content answered by ChatGPT can be applied to my academic learning 3.69 0.928 0.68
5 ChatGPT is a great learning partner. 3.84 0.923 0.72
Avoidance learning motivation
1 I do not like to spend energy on learning tasks. 3.29 1.195 0.80
2 I often think about avoiding learning tasks. 3.33 1.177 0.77
3 I would not want to put in more effort to achieve better learning 3.23 1.150 0.73
performance.
4 I feel like giving up when I need to repeatedly correct my homework. 3.24 1.168 0.73
5 When I do homework, I choose the easiest method possible. 3.47 1.060 0.58
6 I choose to decline when classmates invite me to study together. 3.41 1.197 0.77
7 I feel like giving up when I encounter difficulties in learning. 3.38 1.188 0.80
8 When I don’t understand something in my studies, I don’t tend to actively 3.25 1.090 0.72
seek answers.
ChatGPT dependence
1 I often use ChatGPT involuntarily. 3.86 0.868 0.77
2 I always prefer to use ChatGPT to find information. 3.90 0.837 0.71
3 My life is filled with conversations with ChatGPT. 3.80 0.827 0.73
4 I find it difficult to stop using ChatGPT. 3.74 0.827 0.68
5 ChatGPT is like an intimate friend. 3.77 0.851 0.70
6 I would feel uncomfortable without ChatGPT. 3.81 0.860 0.69

Acknowledgements We are grateful to all researchers who provide data information about their studies
on request.

Author contributions Conceptualization, J.-H.Y., and X.Y.; formal analysis, J.-H.Y.; funding acquisition,
J.-H.Y.; investigation, J.-H.Y., M.Z., and X.Y.; project administration, J.-H.Y., and X.Y.; supervision, X.Y.;
writing—original draft, J.-H.Y., W.N., and L.W.; writing—review and editing, M.Z., and X.Y. All authors
have read and agreed to the published version of the manuscript.

Funding This work was supported by Fundamental Research Funds for the Central Universities in China
(Grant Number: 2022NTSS52), First-Class Education Discipline Development of Beijing Normal Uni-
versity: Excellence Action Project (Grant Number: YLXKPY-XSDW202408) and 2024 Beijing Normal
University’s Teachers’ Teaching Development Fund Project (Grant Number: 2024125).

Data availability The data presented in this study are available on request from the first author.

Declarations

Ethics approval and consent to participate This research obtained approval from the Academic Ethics
Committee of Beijing Normal University. The content of this study was independently written by the
authors, ensuring 100% originality. However, in the “4.5 Discussion” section, answers provided by Chat-
GPT during conversations were used to support the analysis results of this study. These sections have been
duly annotated.

Informed consent All procedures followed were in accordance with the ethical standards of the respon-
sible committee on human experimentation (institutional and national) and with the Helsinki Declaration
of 1975, as revised in 2000 (5). Informed consent was obtained from all patients for being included in the
study.

13
Education and Information Technologies

Conflict of interest The authors have no conflicts of interest to declare.

References
Abedi, G., Rostami, F., & Nadi, A. (2015). Analyzing the dimensions of the quality of life in hepatitis B
patientsusing confirmatory factor analysis. Global Journal of Health Acience, 7(7), 22–31. https://
doi.org/10.5539/gjhs.v7n7p22
Ahmad, S., Zulkurnain, N. N. A., & Khairushalimi, F. I. (2016). Assessing the validity and reliability of
a measurement model in structural equation modeling (SEM). Journal of Advances in Mathematics
and Computer Science, 15(3), 1–8. https://doi.org/10.9734/BJMCS/2016/25183
Ahmad, S. F., Han, H., Alam, M. M., Rehmat, M. K., Irshad, M., Arraño-Muñoz, M., & Ariza-Montes,
A. (2023). Impact of artificial intelligence on human loss in decision making, laziness and safety in
education. Humanities and Social Sciences Communications, 10(1), 1–14. https://doi.org/10.1057/
s41599-023-01787-8
Ali, K., Barhom, N., Tamimi, F., & Duggal, M. (2024). ChatGPT—A double-edged sword for healthcare
education? Implications for assessments of dental students. European Journal of Dental Education,
28(1), 206–211. https://doi.org/10.1111/eje.12937
Baird, A., & Maruping, L. M. (2021). The next generation of research on IS use: A theoretical frame-
work of delegation to and from agentic IS artifacts. MIS Quarterly, 45(1), 315–341. https://doi.
org/10.25300/MISQ/2021/15882
Bouzar, A., Idrissi, E. L., K., & Ghourdou, T. (2024). Chatgpt and academic writing self-efficacy: Unveil-
ing correlations and technological dependency among postgraduate students. Arab World English
Journal, 225–236. https://doi.org/10.24093/awej/ChatGPT.15. 2024(Special Issue on ChatGPT).
Brand, M., Young, K. S., Laier, C., Wölfling, K., & Potenza, M. N. (2016). Integrating psychological
and neurobiological considerations regarding the development and maintenance of specific internet-
use disorders: An Interaction of person-affect-cognition-execution (I-PACE) model. Neuroscience &
Biobehavioral Reviews, 71, 252–266. https://doi.org/10.1016/j.neubiorev.2016.08.033
Brand, M., Wegmann, E., Stark, R., Müller, A., Wölfling, K., Robbins, T. W., & Potenza, M. N. (2019). The
Interaction of person-affect-cognition-execution (I-PACE) model for addictive behaviors: Update,
generalization to addictive behaviors beyond internet-use disorders, and specification of the pro-
cess character of addictive behaviors. Neuroscience & Biobehavioral Reviews, 104, 1–10. https://doi.
org/10.1016/j.neubiorev.2019.06.032
Brandtner, A., Antons, S., Cornil, A., & Brand, M. (in press). Integrating desire thinking into the I-PACE
model: A special focus on internet-use disorders. Current Addiction Reports. https://doi.org/10.1007/
s40429-021-00400-9
Chen, X., Xie, H., Zou, D., & Hwang, G. J. (2020). Application and theory gaps during the rise of artificial
intelligence in education. Computers and Education: Artificial Intelligence, 4, 100002. https://doi.
org/10.1016/j.caeai.2022.100118
Cheung, G. W., Cooper-Thomas, H. D., Lau, R. S., & Wang, L. C. (in press). Reporting reliability, con-
vergent and discriminant validity with structural equation modeling: A review and best-practice rec-
ommendations. Asia Pacific Journal of Management. https://doi.org/10.1007/s10490-023-09871-y
Connaway, L. S., Dickey, T. J., & Radford, M. L. (2011). If it is too inconvenient I’m not going after it:
Convenience as a critical factor in information-seeking behaviors. Library & Information Science
Research, 33(3), 179–190. https://doi.org/10.1016/j.lisr.2010.12.002
Cor, M. K. (2016). Trust me, it is valid: Research validity in pharmacy education research. Currents in
Pharmacy Teaching and Learning, 8(3), 391–400. https://doi.org/10.1016/j.cptl.2016.02.014
Cotton, D. R., Cotton, P. A., & Shipway, J. R. (in press). Chatting and cheating: Ensuring academic integ-
rity in the era of ChatGPT. Innovations in Education and Teaching International. https://doi.org/10.
1080/14703297.2023.2190148
Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., & Wright, R. (2023). So what
ifChatGPTwrote it? Multidisciplinary perspectives on opportunities, challenges and implications of
generative conversational AI for research, practice and policy. International Journal of Information
Management, 71, 102642. https://doi.org/10.1016/j.ijinfomgt.2023.102642

13
 Education and Information Technologies

Elbanna, S., & Armstrong, L. (2024). Exploring the integration of ChatGPT in education: Adapting for
the future. Management & Sustainability: An Arab Review, 3(1), 16–29. https://doi.org/10.1108/
MSAR-03-2023-0016
Elhai, J. D., Yang, H., Dempsey, A. E., & Montag, C. (2020). Rumination and negative smartphone use
expectancies are associated with greater levels of problematic smartphone use: A latent class analy-
sis. Psychiatry Research, 285, 112845. https://doi.org/10.1016/j.psychres.2020.112845
Elliot, A. J. (2006). The hierarchical model of approach-avoidance motivation. Motivation and Emotion,
30, 111–116. https://doi.org/10.1007/s11031-006-9028-7
Green, S. B., & Salkind, N. (2004). Using SPSS for windows and Macintosh: Analyzing and understanding
data (4th ed.). Prentice-Hall.
Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L. (2019). Multivariate data analysis
(8th ed.). Cengage.
Hair Jr, J. F., Hult, G. T. M., Ringle, C. M., Sarstedt, M., Danks, N. P., & Ray, S. (2021). Partial least
squares structural equation modeling (PLS-SEM) using R: A workbook. Springer Nature. https://doi.
org/10.1007/978-3-030-80519-7_1
Hariguna, T., & Ruangkanjanases, A. (2023). Exploring the flexibility and accuracy of sentiment scoring
models through a hybrid KNN-RNN-CNN algorithm and ChatGPT. HighTech and Innovation Jour-
nal, 4(2), 315–326. https://doi.org/10.28991/HIJ-2023-04-02-06
Hong, J. C., Ye, J. H., & Shih, Y. Y. (2019). Positive affect creative self-efficacy on the ability and confi-
dence to predict problem solving avoidance motivation in a digital advertisement design course. Bul-
letin of Educational Psychology, 51(2), 321–339. https://doi.org/10.6251/BEP.201912_51(2).0007
Howard, M. C., & Smith, M. B. (2023). Employee regret and disappointment: Creation of a scale and
foundational application of the approach/avoidance framework. Applied Psychology, 72(2), 419–
450. https://doi.org/10.1111/apps.12367
Hu, B., Mao, Y., & Kim, K. J. (2023). How social anxiety leads to problematic use of conversational
AI: The roles of loneliness, rumination, and mind perception. Computers in Human Behavior, 145,
107760. https://doi.org/10.1016/j.chb.2023.107760
Huang, J., Sun, H., Wang, S., & Zhang, L. (2020). The research progress and prospects of overuse of social
network—based on the I-PACE model perspective. Advances in Psychology, 10, 976–988. https://
doi.org/10.12677/AP.2020.107117
Hwang, M. Y., Hong, J. C., Ye, J. H., Wu, Y. F., Tai, K. H., & Kiu, M. C. (2019). Practicing abductive
reasoning: The correlations between cognitive factors and learning effects. Computers & Education,
138, 33–45. https://doi.org/10.1016/j.compedu.2019.04.014
Javaid, M., Haleem, A., & Singh, R. P. (2023). ChatGPTfor healthcare services: An emerging stage for
an innovative perspective. BenchCouncil Transactions on Benchmarks Standards and Evaluations,
3(1), 100105. https://doi.org/10.1016/j.tbench.2023.100105
Jobst, L. J., Bader, M., & Moshagen, M. (2023). A tutorial on assessing statistical power and determin-
ing sample size for structural equation models. Psychological Methods, 28(1), 207–221. https://doi.
org/10.1037/met0000423
Judijanto, L., Atsani, M. R., & Chadijah, S. (2024). Trends in the development of artificial intelligence-
based technology in education. International Journal of Teaching and Learning, 2(6), 1722–1723.
https://injotel.org/index.php/12/article/view/197
Kahneman, D. (2011). Thinking, fast and slow. Macmillan.
King, M. R., & ChatGPT. (2023). A conversation on artificial intelligence, chatbots, and plagiarism in
higher education. Cellular and Molecular Bioengineering, 16(1), 1–2. https://doi.org/10.1007/
s12195-022-00754-8
Korsgaard, M. A., & Roberson, L. (1995). Procedural justice in performance evaluation: The role of instru-
mental and non-instrumental voice in performance appraisal discussions. Journal of Anagement,
21(4), 657–669. https://doi.org/10.1177/014920639502100404
Kosov, M. E., Malashenko, G. T., Frumina, S. V., Grishina, O. A., Polyakova, O. A., Alandarov, R. A.,
Ponkratov, V. V., Shmigol, N. S., Dzusova, S. S., & Abbood, A. A. A. (2023). Increasing the effec-
tiveness of pedagogical technologies in education: Psychological experience of technological change
management. Emerging Science Journal, 7, 49–63. https://doi.org/10.28991/ESJ-2023-SIED2-05
Ma, X., & Huo, Y. (2023). Are users willing to embrace ChatGPT? Exploring the factors on the acceptance
of chatbots from the perspective of AIDUA framework. Technology in Society, 102362.

13
Education and Information Technologies

Mogavi, R. H., Deng, C., Kim, J. J., Zhou, P., Kwon, Y. D., Metwally, A. H. S., Tlili, A., Bassanelli, S.,
Bucchiarone, A., Gujar, S., Nacke, L. E., & Hui, P. (2024). ChatGPT in education: A blessing or a
curse? A qualitative study exploring early adopters’ utilization and perceptions. Computers in Human
Behavior: Artificial Humans, 2(1), 100027. https://doi.org/10.1016/j.chbah.2023.100027
Morales-García, W. C., Sairitupa-Sanchez, L. Z., Morales-García, S. B., & Morales-García, M. (2024).
Development and validation of a scale for dependence on artificial intelligence in university students.
Frontiers in Education, 9, 1323898. https://doi.org/10.3389/feduc.2024.1323898
Nong, W., He, Z., Ye, J. H., Wu, Y. F., Wu, Y. T., Ye, J. N., & Sun, Y. (2023). The relationship between
short video flow, addiction, serendipity, and achievement motivation among Chinese vocational
school students: The post-epidemic era context. Healthcare, 11(4), 462. https://doi.org/10.3390/
healthcare11040462
Panda, S., & Kaur, N. (2023). Exploring the viability ofChatGPTas an alternative to traditional chatbot sys-
tems in library and information centers. Library Hi Tech News, 40(3), 22–25. https://doi.org/10.1108/
LHTN-02-2023-0032
Park, N., Kim, Y. C., Shon, H. Y., & Shim, H. (2013). Factors influencing smartphone use and depen-
dency in South Korea. Computers in Human Behavior, 29(4), 1763–1770. https://doi.org/10.1016/j.
chb.2013.02.008
Pennycook, G., & Rand, D. G. (2020). Who falls for fake news? The roles of bullshit receptivity, over-
claiming, familiarity, and analytic thinking. Journal of Personality, 88(2), 185–200. https://doi.
org/10.1111/jopy.12476
Pinto, G. A., Vieira, K. C., Carvalho, E. G., & Sugano, J. Y. (2019). Applying the lazy user theory to
understand the motivations for choosing carpooling over public transport. Sustainable Production
and Consumption, 20, 243–252. https://doi.org/10.1016/j.spc.2019.07.002
Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method biases in
behavioral research: A critical review of the literature and recommended remedies. Journal of
Applied Psychology, 88(5), 879–903. https://doi.org/10.1037/0021-9010.88.5.879
Qasem, F. (2023). ChatGPTin scientific and academic research: Future fears and reassurances. Library Hi
Tech News, 40(3), 30–32. https://doi.org/10.1108/LHTN-03-2023-0043
Rodrigues Vasconcelos, M. A., & dos Santos, P. (2023). R. Enhancing STEM learning with Chat-
GPT and Bing Chat as objects-to-think-with: A case study. https://papers.ssrn.com/sol3/papers.
cfm?abstract_id=4434060
Simons, J., Dewitte, S., & Lens, W. (2004). The role of different types of instrumentality in motivation,
study strategies, and performance: Know why you learn, so you’ll know what you learn! British
Journal of Educational Psychology, 74(3), 343–360. https://doi.org/10.1348/0007099041552314
Singharat, W., Kraiwanit, T., & Shaengchart, Y. (2023). The Lazy Economy in a developing country. Cor-
porate & Business Strategy Review, 4(4), 8–15. https://doi.org/10.22495/cbsrv4i4art1
Su, W., Sun, X., Guo, X., Zhang, W., & Li, G. (2022). An analysis of awe evoked by COVID-19 on green
purchasing behavior: A dual-path effect of approach-avoidance motivation. Frontiers in Psychology,
13, 952485. https://doi.org/3389.2022/fpsyg.952485.
Sun, Y., He, J., Li, Y., Yu, L., Li, W., Jin, W., Fu, Y., Zhang, G., & Wang, Q. (in press). Social anxiety
and problematic smartphone use in Chinese college students: The mediating roles of coping style
and the moderating role of perceived friend support. Current Psychology. https://doi.org/10.1007/
s12144-024-05699-x
Tay, S. W., Ryan, P., & Ryan, C. A. (2016). Systems 1 and 2 thinking processes and cognitive reflection
testing in medical students. Canadian Medical Education Journal, 7(2), e97–e103.
Thomas, A. C., Allen, F. L., Phillips, J., & Karantzas, G. (2011). Gaming machine addiction: The role
of avoidance, accessibility and social support. Psychology of Addictive Behaviors, 25(4), 738–744.
https://doi.org/10.1037/a0024865
Tlili, A., Shehata, B., Adarkwah, M. A., Bozkurt, A., Hickey, D. T., Huang, R., & Agyemang, B. (2023).
What if the devil is my guardian angel:ChatGPTas a case study of using chatbots in education. Smart
Learning Environments, 10(1), 15. https://doi.org/10.1186/s40561-023-00237-x
Van Dis, E. A., Bollen, J., Zuidema, W., Van Rooij, R., & Bockting, C. L. (2023). ChatGPT: Five priorities
for research. Nature, 614(7947), 224–226. https://doi.org/10.1038/d41586-023-00288-7
Wang, P., Wang, X., Gao, T., Yuan, X., Xing, Q., Cheng, X., & Tian, M. (2023). Problematic internet use in
early adolescents: Gender and loneliness differences in a latent growth model. Psychology Research
and Behavior Management, 16, 3583–3596. https://doi.org/10.2147/PRBM.S428422

13
 Education and Information Technologies

Wirth, J., Maier, C., Laumer, S., & Weitzel, T. (2022). Laziness as an explanation for the privacy para-
dox: A longitudinal empirical investigation. Internet Research, 32(1), 24–54. https://doi.org/10.1108/
INTR-10-2019-0439
Ye, J. H., Chen, M. Y., & Hao, Y. W. (2023). Teaching and learning in higher education: The role of emotion
and cognition. Frontiers in Psychology, 14, 1230472. https://doi.org/10.3389/fpsyg.2023.1230472
Ye, J. H., He, Z., Yang, X., Lee, Y. S., Nong, W., Ye, J. N., & Wang, C. L. (2023b). Predicting the avoid-
ance learning motivation, learning commitment, and silent classroom behavior of Chinese vocational
college atudents caused by short video addiction. Healthcare, 11(7), 985. https://doi.org/10.3390/
healthcare11070985
Young, K. S., & Brand, M. (2017). Merging theoretical models and therapy approaches in the context
of internet gaming disorder: A personal perspective. Frontiers in Psychology, 8, 1853. https://doi.
org/10.3389/fpsyg.2017.01853
Zhang, S., Zhao, X., Zhou, T., & Kim, J. H. (2024). Do you have AI dependency? The roles of aca-
demic self-efficacy, academic stress, and performance expectations on problematic AI usage behav-
ior. International Journal of Educational Technology in Higher Education, 21(1), 34. https://doi.
org/10.1186/s41239-024-00467-0
Zyphur, M. J., Bonner, C. V., & Tay, L. (2023). Structural equation modeling in organizational research: The
state of our science and some proposals for its future. Annual Review of Organizational Psychology and
Organizational Behavior, 10, 495–517. https://doi.org/10.1146/annurev-orgpsych-041621-031401

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under
a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted
manuscript version of this article is solely governed by the terms of such publishing agreement and appli-
cable law.

Authors and Affiliations

Jian-Hong Ye1,2 · Mengmeng Zhang3 · Weiguaju Nong4 · Li Wang5 ·

Xiantong Yang6

Xiantong Yang
[email protected]
Jian-Hong Ye
[email protected]
Mengmeng Zhang
[email protected]
Weiguaju Nong
[email protected]
Li Wang
[email protected]

1
Faculty of Education, Beijing Normal University, Beijing 100875, China
2
National Institute of Vocational Education, Beijing Normal University, Beijing
100875, China
3
School of Education, Minzu University of China, Beijing 100081, China
4
School of Continuing Education, Hainan Vocational University of Science and Technology,
Haikou 571126, China

13
Education and Information Technologies

5
School of Education, Guangxi University of Foreign Languages, Nanning 530222, China
6
Faculty of Psychology, Beijing Normal University, No. 19 Xinjiekouwai Street,
Beijing 100875, China

13