Linking Neurodiversity and the Adaptation to Emerging Technology: The Interplay between ADHD,Executive Functions,and Chatbot Acceptance

Abstract

This study integrates technology acceptance with adaptation theory to explore the role of cognitive differences, specifically neurodiversity and executive functioning, in the adaptation to the use of chatbots. Using a chatbot providing knowledge on eco-driving, we examine how symptoms of attention deficit hyperactivity disorder (ADHD) and executive functions, such as cognitive flexibility and inhibitory control, influence chatbot acceptance. Findings revealed that ADHD symptomatology did not correlate with cognitive flexibility but was linked to inhibitory control. Cognitive flexibility was the only factor significantly related to technology acceptance in perceived ease of use. However, no significant relationship was found with the intention to use the eco-driving chatbot in mediation analyses. The results highlight the importance of experience with chatbots in mitigating the impact of cognitive flexibility on technology acceptance. This enriches cognitive science and human factors research, offering insights on the role of cognitive flexibility in the perception of usability for inclusive technologies.

Introduction

Emerging technologies such as conversational agents (chatbots) have become vital tools, for example, in health care (Mohamed Jasim et al., 2025) and learning environments (Dai et al., 2024). Despite their growing importance across different sectors, uncertainty remains regarding their appropriate use (Aggarwal et al., 2023). Technology acceptance frameworks (e.g., Davis, 1989; Venkatesh & Davis, 2000) offer valuable insights into decision-making but tend to overlook the role of cognitive differences between users. Meanwhile, adaptation models (e.g., Ployhart & Bliese, 2006; van Dam, 2013) acknowledge such individual differences but have been sparingly applied to human adaptation to emerging technologies (Gößwein & Liebherr, 2024).

Addressing this gap, we integrate technology acceptance with adaptation theory while accentuating neurodiversity and executive functioning. By disentangling the association between cognitive differences and technology adoption, we aim to unravel paths towards enhancing the accessibility of emerging technologies. This contributes to cognitive science, human-computer interaction, and the design of more inclusive technologies.

The technology acceptance model (TAM; Davis, 1989) and subsequent frameworks (Venkatesh et al., 2003, 2012) have been applied to explain the use of chatbots. A current review by Wutz et al. (2023) identified 13 factors that influence chatbot adoption in health care contexts. The most commonly studied identified variables were performance expectancy (henceforward “perceived usefulness,” PU) and effort expectancy (“perceived ease of use,” PEU). Notably, mixed results for both variables were identified. These findings, along with the influence of prior experience with the technology, highlight the need for further research. In our study, the TAM is modeled through the variables PU and PEU of the chatbot, which are hypothesized to influence the intention to use the chatbot (IU). Also, PU is supposed to mediate the effect of PEU on IU.

User characteristics in adaptation to chatbots need further investigation due to the lack of generalizability in prior studies (Aggarwal et al., 2023; Wutz et al., 2023). Our contribution focuses on neurodiversity and executive functions as key cognitive aspects. Neurodiversity is a concept normalizing heterogeneity in neurocognition to foster acceptance of neuro-minorities (Goldberg, 2023). The present study considers attention deficit hyperactivity disorder (ADHD) and its interplay with executive functioning (Barkley & Murphy, 2011; Diamond, 2013). Executive functions have been shown to play a significant role in shaping individuals’ adaptation to technology (Gößwein & Liebherr, 2024). However, the interplay between neurodiversity and the adaptation to chatbots has seldom been considered concerning ADHD (Jang et al., 2021) and, to our knowledge, has never been investigated outside of clinical contexts.

At the same time, a link between ADHD and dangerous driving has been established (Burns et al., 2022; Harzand-Jadidi et al., 2025; Jerome et al., 2006; Sadeghi et al., 2020), potentially putting neurodiverse individuals at risk and polluting the environment due to less sustainable driving behavior. Meanwhile, chatbots offer advantages in delivering cost-effective cognitive behavioral therapy for ADHD (Jang et al., 2021). Still, they can also be used to provide knowledge beyond formal learning environments (Casillo et al., 2022). This is especially relevant as global challenges such as climate change require individual learning and behavioral adaptation (Saran et al., 2024; Zhu et al., 2022). Within the present study, the chatbot transferred knowledge to learners about eco-driving, referring to environmentally friendly driving behavior (Barkenbus, 2010). Eco-driving is supported by emerging technologies, for example, through providing feedback on individual driving behavior (Gimpel et al., 2022; Lin & Wang, 2022; Stephens, 2022).

Our contribution aims to explore the interplay between ADHD, executive functioning, and adaptation to an eco-driving chatbot. Derived from acceptance models, we hypothesize a positive association between PU, PEU, and the intention to use the chatbot. Associations between ADHD scores and inhibitory control, as well as cognitive flexibility, are hypothesized. Following adaptation theory, scoring high for ADHD, as well as deficits in executive functioning, are presumed to be negatively associated with chatbot acceptance. The results could inform the development of chatbots that make eco-driving knowledge accessible to everyone, regardless of their neurodiversity, thereby promoting equal access to technology and reducing social inequality.

Method

The study followed the ethical standards laid down in the Declaration of Helsinki and was approved by the local ethics committee (ID: 2411APGE9634). An online survey was implemented using LimeSurvey; data collection was open from December 2024 to March 2025. Recruitment focused on the platform Survey Circle; the study was titled “Neurodiversity and Chatbots”. Inclusion criteria were being aged above 18 years and speaking either German or English. Statistical analyses were conducted using Jamovi 2.3.21.0. Pearson’s correlations were calculated and interpreted according to Cohen (1988).

One hundred and twenty-six participants completed the survey. Eleven participants failed to give the right answer to the manipulation check, consisting of the question “What is eco-driving?”, after the chatbot interaction. Furthermore, nine answers were excluded because of failing an attention check item. No outliers were detected. The cleaned-up sample consists of N = 106 participants aged between 19 and 57 years (M = 27.65, SD = 6.37; women: 71, men: 33, diverse: 2). The survey mainly reached people with an academic background; 59.4% (63) of the sample reported having a university degree and 37.7% (40) indicating that A-levels were their highest degree.

Participants reported their demographics and experience with chatbots (scores from 1 “never or almost never” to 5 “daily or almost daily”), then filled out a questionnaire on neurodiversity (six-item adult ADHD self-report scale, ASRS, with the sum score calculated following Kessler et al., 2005), and executive functioning (Cognitive Flexibility Inventory, CFI, Dennis & Vander Wal, 2010; Barratt Impulsiveness Scale, BIS-15, Meule et al., 2020). The survey continued with an online version of Berg’s Card Sorting Task (BCST) to assess cognitive flexibility objectively (Vékony, 2022). In the following, we distinguish the measurements by naming the CFI score “subjective” and the number of perseverative errors in the BCST “objective” cognitive flexibility. Please note that a higher number of perseverative errors points towards lower objective cognitive flexibility.

Finally, participants interacted with the eco-driving chatbot, which was based on predefined question-and-answer choices, thereby ensuring comparability of the experience across participants. The chatbot was implemented through the platform Botpress and provided basic knowledge about eco-driving; its acceptance was assessed with a questionnaire based on Davis (1989) and Agarwal and Karahanna (2000), ranging from 1 to 7. All implemented scales demonstrated acceptable to excellent internal validity (George & Mallery, 2003); Cronbach’s alpha values ranged from .771 (ASRS) to .953 (IU).

Results

Descriptives

Chatbot acceptance was generally high, M_PEU = 6.39 (SD = 0.87), M_PU = 5.27 (SD = 1.20), M_IU = 4.98 (SD = 1.52), as well as previous experience with chatbots: 60.4% of the participants (64) reported using chatbots at least once per week. 23.6% of the participants (25) scored in the highest stratum on the ASRS, so positive for ADHD. Meanwhile, subjective cognitive flexibility (M = 5.42, SD = 0.67) and impulsivity (M = 34.05, SD = 6.30) were above scale averages with a notably low standard deviation for cognitive flexibility. Perseverative errors showed a range of 3 to 16 errors, but a relatively low standard deviation (M = 7.40, SD = 2.41), indicating a low variance in objective cognitive flexibility.

Further Analyses

Chatbot acceptance

The correlation results reflect the hypothesized associations between PU, PEU, and IU with significant, moderate to strong correlations (r_PU|IU = .582, r_PEU|IU = .447, r_PU|PEU = .672, all p < .001). Mediation analysis using bootstrapping with 10,000 samples revealed a significant link between the predictor PEU and the mediator PU (B = 0.924, p < .001), and a significant link between PU and IU (B = 0.651, p < .001), while the direct path between PEU and IU turned out to be non-significant (B = 0.177, p = .355). After entering PU into the model, PEU predicted IU significantly, B = 0.778, p < .001.

Neurodiversity and Executive Functioning

ADHD scores did not significantly correlate with subjective cognitive flexibility (r = −.029, p = .770) or objective cognitive flexibility (r = −.082, p = .402), contradicting our presumptions. In line with our presumptions, ADHD scores correlated significantly with inhibitory control (r = .454, p < .001). Building on this pattern of results, we consider ADHD, cognitive flexibility (subjective and objective), and inhibitory control separately in the following analyses.

Chatbot Acceptance and Cognitive Differences

The correlation matrix revealed two significant links for chatbot acceptance and cognitive differences: between subjective cognitive flexibility and PEU (r = .205, p = .035) and between objective cognitive flexibility and PEU (r = −.267, p = .006); the other associations were non-significant and are therefore not explored further.

A multiple regression model was calculated to test whether the two assessments of cognitive flexibility would be predictors of PEU. Indeed, the results indicated that these two predictors explained 10.9% of the variance in PEU (R² = .109, F(2,103) = 6.30, p = .003). It was found that subjective cognitive flexibility significantly predicted PEU (β = .195, p = .039), as well as objective cognitive flexibility (β = −.259, p = .006).

Using subjective cognitive flexibility as the predictor, a mediation model using bootstrapping with 10,000 samples revealed a non-significant direct path between cognitive flexibility and IU (B = −0.128, p = .540). The path between this measurement of cognitive flexibility and the mediator PEU edged towards significance (B = 0.266, p = .083) but was non-significant, nevertheless. After entering PEU into the model, the direct link between subjective cognitive flexibility and IU remained non-significant (B = 0.084, p = .707); the indirect path through PEU approached significance (B = 0.212, p = .083).

When calculating the mediation model (bootstrapping, 10,000 samples) with objective cognitive flexibility as the predictor, the direct path between the predictor and IU was significant (B = 0.095, p = .019). The path between cognitive flexibility and PEU as the mediator, however, was non-significant (B = −0.077, p = .130). After entering PEU into the model, the direct link from objective cognitive flexibility to IU (B = 0.032, p = .589) and the indirect path through PEU (B = −0.064, p = .154) turned out to be non-significant.

Discussion

Our findings indicate that chatbot acceptance can be modeled as suggested by the TAM (e.g., Davis, 1989; Venkatesh & Davis, 2000), but the non-significant direct path between PEU and IU has to be noted. This result aligns with prior research, which has demonstrated a negative effect of restrictive user input on PEU (Wutz et al., 2023).

The absence of a relationship between ADHD symptoms and both assessments of cognitive flexibility is consistent with recent findings questioning the conventional link between ADHD and cognitive flexibility deficits (Aydın et al., 2022; Li et al., 2023). It should be noted that the ASRS (Kessler et al., 2005) is a screening tool for ADHD rather than being a formal diagnostic instrument. In the current study, it was administered to a community sample rather than to clinically diagnosed individuals.

However, cognitive flexibility was found to predict PEU and is therefore connected to chatbot acceptance. Still, the findings on (subjective and objective) cognitive flexibility contradict our presumptions. This might be explained by the missing link within chatbot acceptance and by specific characteristics of our sample. The high level of prior experience with the technology might have had an impact, as participants were (partially) adapted to chatbots, therefore requiring less cognitive flexibility and mitigating the association between inhibitory control and technology acceptance. Also, our sample lacked variance in cognitive flexibility, which could diminish existing effects.

While the novel topic is a strength of this study, like every empirical work, it does not come without limitations. Despite its increasing media presence and rising diagnosis rates, ADHD is relatively rare in Germany, affecting approximately 3% of adults aged 18 to 69, with 0.4% receiving a formal diagnosis (Bachmann et al., 2017; de Zwaan et al., 2012). Our sample is likely skewed due to self-selection bias, which may have influenced the characteristics of the individuals who participated in the study. However, this provided the needed variance to analyze the relationship between neurodiversity and the adaptation to chatbots.

Furthermore, the chatbot interaction was rather restricted as a result of the use of pre-defined questions and answers. As chatbots are becoming increasingly widespread, the chatbot configuration might have been perceived as a technological regression instead of an emerging technology. Although we purposefully chose such a configuration to ensure comparability of participant experience, future research should look into the acceptance of more open chatbots, for example, based on a large language model.

Our study successfully explored the interplay between ADHD, executive functioning, and adaptation to an eco-driving chatbot. We did not find an association between neurodiversity and the intention to use the chatbot, thereby highlighting the value of chatbots as a tool for inclusive knowledge transfer. Cognitive flexibility and its connection to usability should, however, be considered in designing those technologies, for example, by promoting new applications to cognitively flexible individuals first.

Conclusion

The role of cognitive differences in adapting to emerging technologies was examined in an academically well-qualified and technically experienced sample using a chatbot, which provided basic knowledge about eco-driving. Our findings challenged the conventional link between ADHD and cognitive flexibility deficits but supported the link to inhibitory control. Surprisingly, only cognitive flexibility was correlated with technology acceptance, but the effect was not found in a subsequent mediation analysis. These insights suggest that experience with chatbots decreases the need for cognitive flexibility and mitigates the relation between inhibitory control and technology acceptance, thereby supporting prior findings on executive function demands being highest in a novel task (Phillips, 1997; Shallice & Burgess, 1991). High levels of and low variance in cognitive flexibility might have affected the findings.

Lessons learned emphasize the need for a nuanced understanding of technological adaptation that considers neurodiversity. Future studies should further investigate the interplay of cognitive differences and the adaptation to new technologies, which could foster inclusive technology design and provide impacts at the intersection of cognitive science, human-computer interaction, and the implementation of emerging sustainable technologies.