1. Introduction
Cognitive flexibility, which is characterized by the capacity to shift between etic and emic perspectives, is mentioned across various definitions and models of intercultural competence (Aski et al., Reference Aski, Jiang and Weintritt2023; Byram & Wagner, Reference Byram and Wagner2018; Deardorff, Reference Deardorff2006; Fantini, Reference Fantini and Deardorff2009), highlighting its significance in “developing the capacity to ‘move between’ linguistic, cultural and knowledge systems” (Scarino & Liddicoat, Reference Scarino and Liddicoat2016, p. 21). In Deardorff’s (Reference Deardorff2006) model, flexibility is positioned as a desired internal outcome that emerges through the process of developing intercultural competence. It functions as both an internal orientation and a readiness to consider alternative perspectives by actively reorganizing one’s meaning framework and responding to dissonance or difference by adapting interpretations of meaning across linguistic and cultural contexts. Framing cognitive flexibility as an active disposition underscores both its enabling role in the development of intercultural competence (O’Dowd & Dooly, Reference O’Dowd, Dooly and Jackson2020) and its inherently subjective, dynamic nature. A central process in the development of this disposition is decentering: stepping “outside one’s existing, culturally constructed, framework of interpretation” (Liddicoat & Scarino, Reference Liddicoat and Scarino2013, p. 58) to recognize the legitimacy of alternative worldviews. Through exposure to new interpretive frames (Álvarez Valencia & Michelson, Reference Álvarez Valencia and Michelson2023; Drewelow, Reference Drewelow2025), one can become aware of when one’s familiar meaning-making resources may be partial, limited, or culturally situated. This decentering is a necessary part of the perspective shifting and meaning frame reorganization that underpin cognitive flexibility.
This study aims to contribute to the growing body of research on second language (L2) pedagogies focused on the process of developing interculturality (Álvarez Valencia & Michelson, Reference Álvarez Valencia and Michelson2023; Aski et al., Reference Aski, Jiang and Weintritt2023; Drewelow, Reference Drewelow2023; Lomicka & Ducate, Reference Lomicka and Ducate2021) by exploring whether interacting with ChatGPT (OpenAI, 2024) offers a context that may be conducive to expanding students’ interpretive frameworks and enabling shifting between perspectives. As an artificial intelligence (AI)–powered language model designed to generate human-like responses, ChatGPT can produce varied, sometimes unexpected, responses to a single prompt. This study investigates whether ChatGPT’s responses challenged students’ existing assumptions and interpretations and prompted them to engage in the reflective process of comparing, questioning, and reassessing to consider alternative cultural framings.
The study’s data come from a university-level advanced Language for Specific Purposes (LSP) French course focused on marketing and advertising in France at a large R1 university in the southeastern United States. This course is designed as a semester-long immersive learning experience, with learning by action and authenticity as governing principles, and is informed by Kolb’s (Reference Kolb1984) model of experiential learning. Over 14 of the 16 weeks in the semester, students participated in a simulation designed to guide them in developing a product or service for the French consumer market. In-class activities and assignments provided a structured environment that helped students define their product or service specifications, positioning, and messaging. The integration of ChatGPT into the simulation offered an additional opportunity to gather insights into consumer needs and preferences in France, while also supporting idea generation, strategic thinking, and decision-making.
2. Literature review
2.1. Moving beyond familiarity: Cultivating cognitive flexibility
Developing the capacity to shift between interpretive frameworks involves recognizing how meaning might change when viewed from another interpretive position. As Byram and Cain (Reference Byram, Cain, Byram and Fleming1998) observe, individuals’ familiar values, beliefs, assumptions, and cultural references guide their interpretation of difference and can “interfere with [the] perception of other cultural systems” (p. 36). Applying one’s own cultural frames of reference to interpret communications, situations, or issues from another culture is difficult to evade because, as Agar (Reference Agar1994) notes, “we see and hear and otherwise experience very largely as we do because the language habits of our community predispose certain choices of interpretation” (p. 66). Moving beyond an intracultural viewpoint toward an intercultural stance therefore requires more than exposure to difference; it requires the capacity to suspend the automatic recourse to familiar meaning-making resources.
However, the readiness to interpret meaning through cultural perspectives different from one’s own should not be assumed to arise automatically in L2 learning. As Johnson (Reference Johnson2015) explains, when students encounter unfamiliar worldviews, they may discard the information, absorb it into their existing worldview, or experience dissonance between prior interpretations and new information. For her, it is through intentional engagement with this dissonance that a shift in understanding can emerge. This emphasis on engaging with dissonance underscores the role of the learning environment in supporting the reorganization of students’ interpretive resources and helping them learn to see the unfamiliar from an insider’s viewpoint and the familiar from an outsider’s.
2.2. Pedagogical conditions for developing cognitive flexibility
Creating conditions for perspective shifting requires exposing students to viewpoints informed by alternative meaning-making resources (values, assumptions, and cultural references) that can disrupt the familiar. Introducing interpretive dissonance or consonance can help students notice how the interpretation of meaning depends on cultural positioning and subjective experience (Drewelow, Reference Drewelow2023, Reference Drewelow2025; Drewelow & Finney, Reference Drewelow and Finney2020; Liddicoat & Scarino, Reference Liddicoat and Scarino2013; Zarate, Reference Zarate1993). Discrepancies render visible cultural references that may not have been previously considered, creating opportunities to question assumptions and connect with the subjectivity of others. Similarities invite reflections on one’s own cultural positioning and how it connects with others’ intraculturality and lived experiences.
Encounters with multiple frames of interpretation should be supported by activities that involve cognitively processing “the asymmetrical imbalance between the familiar and the strange” (Matos, Reference Matos, Fäcke, Gao and Garrett-Rucks2025, p. 266), such as comparing, questioning, and experimenting with interpretations from perspectives other than one’s own (Andreotti & de Souza, Reference Andreotti and de Souza2008). Acknowledging this imbalance and reflecting on the potential dissonance it generates can allow new perceptions to emerge and contribute to decentering from taken-for-granted cultural frames (Matos, Reference Matos, Fäcke, Gao and Garrett-Rucks2025). Supporting the development of cognitive flexibility in the L2 classroom thus requires contexts that help students practice stepping back from initial interpretations (Drewelow, Reference Drewelow2025), entertain multiple possible interpretations (Álvarez Valencia & Michelson, Reference Álvarez Valencia and Michelson2023), and cultivate tolerance for ambiguity by staying engaged with dissonance and uncertainty (Matos, Reference Matos, Fäcke, Gao and Garrett-Rucks2025).
2.3. ChatGPT as a context for perspective shifting
Although generative AI does not constitute a cultural “other,” its responses may diverge from student expectations, present unfamiliar framings, or introduce alternative interpretations, creating dissonance that can support perspective shifting and cognitive flexibility. A potential caveat is the variability in accuracy and reliability of ChatGPT’s responses (Xiao & Zhi, Reference Xiao and Zhi2023), which can vary depending on prompt wording and the language used, with greater fluctuations in languages other than English (Kohnke et al., Reference Kohnke, Moorhouse and Zou2023). In a study by Hellmich et al. (Reference Hellmich, Vinall, Brandt, Chen and Sparks2024), participants expressed distrust toward ChatGPT due to concerns about the “originality, purity, and veracity” (p. 13) of the information. This skepticism contributed to participants’ choice not to use ChatGPT for L2 learning. However, Xiao and Zhi (Reference Xiao and Zhi2023) found that students “did not blindly accept the information generated by ChatGPT but instead evaluated its accuracy, relevance, and specificity” (p. 12). Some participants in Hellmich et al.’s study also recognized the need to verify the veracity of the information, demonstrating an awareness of ChatGPT’s limitations as a knowledge source. Such evaluative engagement might itself serve as a site for cognitive flexibility: noticing discrepancies between students’ expectations and the AI’s responses may lead them to reassess assumptions, consider alternative perspectives, or reflect on how meaning is shaped by contextual cues. These findings suggest that dialogic interactions with ChatGPT might provide a context for perspective shifting, not because ChatGPT offers culturally “authentic” viewpoints, but because the interpretive disruptions it can introduce might prompt students to compare, question, and reinterpret their own initial understandings.
3. Methods
A qualitative approach was deemed most appropriate to gain a comprehensive understanding of participants’ subjective experiences with ChatGPT. Such an approach allowed the study to capture how participants compared, questioned, and reconsidered their assumptions and perspectives. The following research question guided the study: How did students’ interactions with ChatGPT create conditions that may have supported perspective shifting and cognitive flexibility?
3.1. Use of AI in the course
The course curriculum was designed with a series of sequenced tasks aimed at helping students create marketing-related materials, including a website, a static web advertisement, and a 45-second video advertisement to introduce the product or service they conceptualized. From Weeks 2 through 7, the focus was on studying the French consumer market and defining the product or service concept, with the goal of producing a comprehensive marketing action plan (due Week 9) that served as the foundation for the website development. Integrating ChatGPT at the concept development stage was a strategic decision, as it provided an additional tool to enhance market research and help students make informed decisions early in the creative process. By interacting with AI, students could gather additional insights into consumer preferences, behaviors, and trends from a different source, which supported their ability to refine and develop their product or service concept.
The course met twice a week in face-to-face sessions of 75 minutes. In one class session during Week 4 of the semester, students explored how to use ChatGPT to gather information on economic and social trends in the French consumer market in 2025. To prepare for the class session, the instructor asked ChatGPT in French (version: free GPT-4o) to generate prompts or questions to identify trending market sectors in France, which were then displayed on a Padlet in class. Following a think-pair-share format, students first discussed the questions generated by ChatGPT and debated which ones might yield the most useful insights. Next, each pair was assigned one question from the list and instructed to use ChatGPT (version: free GPT-4o) and a search engine to identify four trending sectors they were interested in. All interactions with ChatGPT, including follow-up prompts, were to be conducted in French. Each pair shared their findings on the Padlet. Finally, a whole-class roundtable discussion followed, where the results from the research were shared, and students discussed their experience using ChatGPT. During Weeks 5 and 6, students were introduced to two key marketing concepts: the 4Ps of marketing (product, price, place, promotion) and SWOT analysis (strengths, weaknesses, opportunities, threats). Over the course of four class sessions, students were given 20 minutes each session to collaborate with their partner on their project. For this task, they were instructed to use ChatGPT in French and a search engine to develop their product or service concept, employing the tools as they saw fit. This involved gathering information on French consumer wants, needs, desires, and behaviors, as well as defining the names of their company and product. Students also researched existing competitive products and companies, determined their product or service pricing, and developed a distribution strategy. During Week 7, they were asked to continue their research outside of class and were encouraged to use ChatGPT in French to further refine their product or service concept. Students were not directly observed while using ChatGPT, and transcripts of their interactions with the tool were not collected.
3.2. Participants
Fourteen students were enrolled in the course, and all were invited to participate in the study. In Week 5 of the semester, a research assistant, who was not involved in teaching or evaluating the course, presented the study at the beginning of a class session using a script provided by the researcher. The students were then invited to sign up for a time slot to participate in the study during either Week 7 or 8 of the semester. They were asked for their consent to use their responses for research purposes. The researcher, who was also the course instructor, was not in the classroom when consent to participate was sought. Students were informed that neither participation nor non-participation would affect their course grades or their relationship with their instructor. Consent forms were kept in the main departmental office and were not released to the researcher/instructor until after final grades had been submitted. All 14 students (12 female, two male) enrolled in the course gave consent to participate in the study. Thirteen participants were native English speakers (12 Americans and one Canadian). One participant was a native Spanish speaker from Colombia. They were between 19 and 23 years old. Four participants were French majors and 10 were minoring in French. L2 proficiency was not directly measured; however, in upper-division courses, enrollment typically consists of students whose proficiency falls within the B1–B2 range of the Common European Framework of Reference for Languages (CEFR; Council of Europe, 2001).
3.3. Data collection and instrument
Data were collected through semi-structured interviews lasting 20–25 minutes. The interviews were a required component of the course, graded as either completed or not completed, and accounted for 5% of the final grade. The semi-structured format allowed for flexibility in responses, with a set of predetermined questions designed to encourage open-ended discussion while enabling comparisons of responses across participants. To help participants focus on processing and connecting with their experiences, the interviews were conducted in English, the native language of the majority of students and the primary language used at the university where the study took place. Although the students were advanced learners of French, expressing their thoughts in French might have hindered their ability to articulate their ideas fully; English was therefore considered the most appropriate language for the interviews.
The interview protocol consisted of seven open-ended questions. Two questions addressed students’ general experience using ChatGPT for research and concept development, asking how helpful they found the tool compared to a search engine, and what challenges or limitations they encountered. Two questions focused on perspective shifting, asking whether and why using ChatGPT challenged their understanding of the French market and its consumers, and whether it prompted them to adjust or change their initial product or service idea. Two questions explored cognitive flexibility, asking whether ChatGPT made it easier to shift between their own point of view and that of someone trying to sell a product in the French market, and how using ChatGPT influenced their ability to think about problems or ideas from multiple perspectives. The final question served as a synthesis, asking what students had found most helpful in the course so far for defining the product or service for their website project.
The individual interviews took place outside of class in a private office on campus, with only the research assistant and the student present. The research assistant was instructed to maintain a neutral, non-judgmental demeanor throughout the interview to ensure the student felt comfortable sharing their thoughts. The research assistant occasionally asked for clarification and encouraged elaboration by requesting examples. The interviews were recorded only if the student had given consent to participate in the study.
3.4. Data analysis procedures
The audio files of each interview were uploaded to Pinpoint (Google, 2025), a free AI-powered tool that can automatically transcribe audio and video files into text. After transcription, the research assistant systematically compared each transcript with the corresponding audio recording and manually made any necessary adjustments to ensure fidelity and accuracy to the original recording. The researcher began data analysis only after final grades were submitted at the end of the semester. Because cognitive flexibility and perspective shifting are subjective, internally experienced processes that unfold dynamically across interactions, the analysis aimed to capture their context-dependent expression in participants’ accounts. Interview transcripts were analyzed using an interpretive, inductive thematic approach (Braun & Clarke, Reference Braun and Clarke2006) to examine both what students reportedly experienced in their interactions with ChatGPT and how they interpreted those experiences in relation to perspective shifting and cognitive flexibility.
Initial categories were identified and iteratively refined through multiple readings of the transcripts, with attention to how students described their interactions with ChatGPT and how these interactions stimulated reflection on existing assumptions and perspectives. The following categories emerged during analysis: distrust; beliefs about ChatGPT; evaluations of ChatGPT’s responses; uncertainty; challenges with ChatGPT; role-playing; making comparisons; curiosity and active questioning; movement between interpretive frameworks; recognition of the limits of their own perspective; consideration of contrasting viewpoints; reflecting across perspectives; and refined understanding. These categories were subsequently consolidated into themes through the researcher’s interpretive engagement with the data and comparison across participants to identify recurring conditions that appeared to support perspective shifting and cognitive flexibility.
Themes were developed inductively and are illustrated through representative excerpts. These excerpts are not intended to be exhaustive. They were selected to reflect patterns observed across the interview dataset rather than isolated instances. Analytical rigor was supported through sustained and iterative engagement with the data, with transcripts read multiple times. Theme development proceeded recursively rather than linearly, allowing earlier interpretations to be revisited and refined. Credibility was supported by grounding analytic claims in participants’ own words, and consistent analytic procedures were applied across all transcripts to support dependability, in line with qualitative methodological principles (Lew et al., Reference Lew, Yang, Harklau, Phakiti, De Costa, Plonsky and Starfield2018).
4. Findings
Considering the co-constructed nature of the interviews, participants’ responses represent their individual perspectives and subjective understanding of their experiences using ChatGPT. The researcher acknowledges that their own subjectivity and interpretive choices influenced the analysis and presentation of the findings.
4.1. Perspective shifting with ChatGPT? It’s complicated
Students’ comments reflect concerns about using ChatGPT to gather information on market trends and consumer needs and preferences in France. They expressed skepticism about its legitimacy and credibility, as well as doubts about its authenticity and cultural positioning. The data point to ChatGPT’s limitations in creating dissonance and supporting interpretive disruptions.
4.1.1. Doubts about ChatGPT’s legitimacy
A dominant preoccupation for students was the credibility of ChatGPT’s responses: whether the information provided was valid and reliable enough to be trusted. In the comment that follows, the student’s skepticism stemmed from her doubts about ChatGPT’s expertise across cultural contexts (“Maybe it doesn’t have all the information”) and its perceived limited ability to provide depth and cultural framing (“doesn’t really tell all the background and stuff of things”). The incomplete or insufficient contextualization introduced uncertainty, leading them to question ChatGPT’s legitimacy as a source of culturally situated knowledge.
With ChatGPT, I feel it’s not as trustworthy with French information that it is with English. Because it’s kind of like you don’t know for sure. […] Maybe it doesn’t have all the information. I just feel ChatGPT sometimes just gives a brief example of something, doesn’t really tell all the background and stuff of things. So, I feel you could say tell me more, but I feel it never tells you everything. (Student 7)
Uncertainty about the source of information contributed to a perceived lack of transparency in ChatGPT. The following comments illustrate the disorientation students experienced when interacting with the AI tool. They struggled to formulate questions that would yield meaningful information and were aware that ChatGPT could produce overgeneralizations or stereotypes, which conflicted with their expectations of factual accuracy and prompted doubts about the legitimacy and trustworthiness of the information.
I wouldn’t necessarily know if the information they’re [ChatGPT] giving me is right or just maybe based on stereotypes or maybe it’s not the actual information. […] I don’t trust it since I don’t know the sources. It will just make me question, is this actual factual information? (Student 10)
The challenges I’ve faced are just kind of knowing what to ask. […] It’s foreign to me so I didn’t quite know what to ask and then wondering, does this already exist? (Student 5)
Doubts about ChatGPT’s credibility were compounded by students’ prior experiences with it in other settings and for other purposes. They recognized “it can give false information or make stuff up” (Student 8). A recurring concern was their ability to “spot” when information was “wrong.” Students felt more confident assessing ChatGPT’s responses on topics close to their cultural background when they could draw on their personal knowledge and cultural references to evaluate validity, accuracy, and acceptability. However, “asking questions about a culture that I don’t exist in” was disorienting, because students felt they lacked the cultural legitimacy (“I’m not French”) to assess accuracy or identify gaps in their own knowledge.
I know that it [ChatGPT] can be wrong and that’s a lot harder to spot in French. So, I normally take it with a grain of salt. [with] my familiarity with the language and also the culture, […] if it’ll tell me something like “all Americans do this” or “all people who go to the University of [XXX] do this” I know it’s wrong because I go here and I’m an American but I’m not French, so I won’t be able to know the cultural difference. (Student 13)
I just feel when I’m using it for more American things, I’ll just know right away if it’s true or not. I feel red flags will go off in my head. […] I’m asking questions about a culture that I don’t exist in, I feel I have to fact-check more because I don’t know. I have no idea if this could be completely wrong and I would have no clue. (Student 14)
These comments suggest that students’ ability to critically evaluate ChatGPT’s responses depends on their cultural familiarity with the topic. As the following comments show, when that familiarity is limited or absent, students may engage with ChatGPT-generated content more superficially, take it less seriously, or even dismiss the information entirely, thereby limiting their ability to meaningfully engage with the content.
Relying on something that I don’t honestly fully trust to tell me the truth every time seems kind of silly. (Student 1)
I still just don’t really trust anything it says that I don’t already know. So, if it says something else, I’m just going to kind of ignore that or I’m going to go separately look it up. (Student 2)
4.1.2. Issues with ChatGPT’s cultural positionality
Students’ perceptions that ChatGPT’s cultural framing was limited led them to question its ability to offer diverse and culturally situated viewpoints. One student noted that its responses felt generalized or “average,” lacking the nuance or critical insight needed to reveal unfamiliar perspectives. This perceived lack of depth contributed to the impression that ChatGPT produced a homogenized cultural narrative, offering consensus rather than culturally diverse perspectives. The student’s observation that it “doesn’t have that many interesting outlying opinions” points to ChatGPT’s limitation in creating a context capable of generating cultural dissonances through alternative perspectives.
I know it [ChatGPT] does have multiple perspectives in itself because it’s pulling stuff from all over, but when it presents it, it sounds like it all seems to be coming from the same place. It usually only gives one answer. I’m sure if you ask for multiple, it might give more. But I feel like ChatGPT in general it’s kind of an average of everything out there. It’s kind of an average opinion, not anything that’s interesting. It doesn’t have that many interesting outlying opinions or anything like that. (Student 3)
Two students viewed ChatGPT as rooted within an American framework. They considered that its responses, shaped by the “United States internet,” reflected an American worldview rather than culturally situated or localized perspectives. Their interactions with ChatGPT were therefore more akin to intracultural than intercultural encounters. One student pointed out that a key challenge with ChatGPT is formulating prompts or questions that generate multiculturally diverse viewpoints or invite perspectives sufficiently unfamiliar to elicit comparisons and prompt reflections on existing assumptions.
ChatGPT is not a French person. ChatGPT was developed by Americans. They use American data. It uses your data to tell you what you want to know, I mean it’s an amalgamation of all of the information on the internet, but particularly the United States internet. (Student 11)
I will say I think that AI is also very American still. So, a lot of times it’ll tell us things about the American market and then you have to redirect it to the French market. Maybe it’s the questions I’m asking ChatGPT, but it doesn’t seem to like challenge anything. I mean I definitely think it’s pretty good at reading what I want the answer to be and giving me that answer. (Student 1)
This comment suggests not only that ChatGPT’s intracultural positioning limits the possibility of encountering alternative perspectives but also that its tendency to adapt its responses to users’ expectations might reinforce preexisting cultural assumptions. The comment below illustrates how a student, in seeking more contextual framing, might formulate prompts and questions to get information that fits their worldview, thus short-circuiting opportunities for cultural dissonance.
If I’m doing research, I want to actually understand and not just know the facts. I can bullet point it, but can I explain it? No. So, that’s the thing, I will just have to keep prompting it [ChatGPT] to give me the answer, which will be me looking for the information that I have in my brain. Maybe it’s like I know that you can give me that answer. So, I’m going to keep prompting you so you can give me the answer I want. (Student 10)
For some students, ChatGPT was accommodating to the point of being manipulative, tailoring its responses to fit within the user’s cultural background to avoid conflict or discomfort for the user. Through their comments, students expressed skepticism about ChatGPT’s ability to challenge existing assumptions or accurately represent alternative viewpoints and viewed it as responsive (shaping its responses to please them) rather than disruptive.
It’s just trying its best. It just wants to make you happy, but I’m like, no, I want the correct answer. It’s giving me what I want to hear. It’s giving you what you want to hear. (Student 2)
I feel because it’s giving you exactly what you’re asking for, a lot of times it is catering to you, it’s not necessarily going to go against what you’re asking for. (Student 8)
I don’t think it really does a good job with trying to change your perspective, but more so with catering to your like needs and your questions and what you want. (Student 12)
4.2. Developing cognitive flexibility with ChatGPT
Students’ comments revealed that interacting with ChatGPT facilitated subjective positioning and stimulated dialectical thinking. The data suggest that ChatGPT enabled conditions that supported the development of cognitive flexibility by helping students recognize the limits of their own perspectives, consider contrasting viewpoints, and reflect across those perspectives to refine understanding.
4.2.1. Enable subjective positioning
Interacting with ChatGPT provided two students with the opportunity to move from abstract knowledge to actively experimenting with acting as if they were insiders (e.g. imagining themselves as a French marketer or product designer). This subjective positioning emerged from the interactive and conversational aspects of the tool, which created a participatory, experiential environment. Dialoguing with ChatGPT simulated a real-world situation that engaged the students’ imagination enough for them to project themselves into plausible roles within a culturally unfamiliar context.
In the following comment, the student co-constructed their understanding of the French market by proposing scenario-based prompts to ChatGPT. The interaction scaffolded a situated understanding and enabled them to adopt an emic perspective. Within that environment, they were able to practice cognitive flexibility. Engaging in scenario-based dialogue with ChatGPT made it easier to “switch back and forth” between interpretive frameworks, because they could experiment with alternate subjective positions.
You can even ask it [ChatGPT] to be in the point of view of anything, so I could ask it how would you approach selling XYZ in the French market or how would you approach marketing XYZ to French consumers and give it different scenarios, so I feel that makes it very easy for me to kind of switch back and forth and also place myself in the position of someone selling to the French market. Not necessarily putting myself in the exact position but gathering the information so that I can act as if I was doing. (Student 6)
By framing their prompts in the first person, another student constructed a scenario in which they could dialogue with ChatGPT from within the imagined role of a product designer. By engaging with the AI tool as if they were actually preparing a launch and expecting feedback, they shifted from an etic perspective (“I am doing a project for a class”) to an emic one (“I am trying to sell this product”), applying a framework different from her own.
When I’m using ChatGPT, I have to write out very much what I’m looking for. […] I have to be “I am doing” … “I am making a product …, it is about …” “this is some of the stuff about my product” I usually describe it as if I am making the product. I’m not like “I am doing a project for a class I need to do this.” I’m saying “I am trying to sell this product.” So, it’s like putting myself more directly in the role just by having to describe it to the chatbot. It’s like an interaction where I’m actually in that role. Someone’s talking back to me as if I actually am doing it. (Student 3)
One student recognized that stepping outside their own perspectives was not a natural inclination. They found that interacting with ChatGPT helped them begin to access and compare alternative cultural viewpoints. Unlike Students 3 and 6, they did not explicitly describe taking on an emic perspective themselves or imagining themselves in a role. However, their comment suggests that ChatGPT’s ability to generate responses tailored to a specific cultural context offered them an entry point into subjective positioning. By asking ChatGPT “what are they looking for?” they were able to compare that information with their own perspectives and build a more situated understanding of the French market, giving them a foundation to take on an emic perspective.
I think that is one of the things I struggle with is kind of getting out of my own perspective. So, thinking in someone else’s perspective is a little bit harder for me. ChatGPT definitely gives you ideas, you can ask for multiple perspectives. Like that market, what are they looking for? It’s spitting out information that’s going to be specifically for a French market. I think that’s just what has really helped me understand the multiple perspectives thing. Cause I obviously have my own already. So, it was helpful to be able to compare the information I was getting from ChatGPT and just creating a more well-rounded idea of what the French market wants compared to me as an American. (Student 4)
4.2.2. Stimulate dialectical thinking
When students used ChatGPT to research French consumer behavior and the French market to refine their product or service idea, the tool’s responses didn’t always align with their prior knowledge, assumptions, or expectations. Experiencing dissonance or discrepancies prompted students to notice subjective positioning and the cultural embeddedness of perspectives, and to recognize the importance of thinking more dialectically in intercultural contexts.
Despite their skepticism about the trustworthiness of ChatGPT’s responses, one student noted that, by exposing alternative viewpoints, their interactions with ChatGPT underlined the necessity of perspective shifting in intercultural contexts. Their comment shows that even when doubting the veracity of ChatGPT’s information (“you might not fully believe”), awareness that a different interpretive framework is in play (“okay, they think differently”) can serve as a trigger to explore dissonance.
It’s nice to have ChatGPT to kind of also just remind you not always intentionally but just remind you that there are those differences culturally and that you do have to shift your perspective. I think that despite a general distrust of how true the information is, I think it still promotes the ability to see different perspectives. So, while you might not fully believe that that is what French people think, it’ll start promoting that, “okay, they think differently” at least and then it’ll promote further research. (Student 13)
When ChatGPT revealed information that they had not previously encountered, one student initially relied on their intracultural knowledge (“my American consumer schemas”) to interpret the new information. Their comment points to the beginning of a shift toward a more dialectical stance, as they attempted to connect these new perspectives with their existing ones and develop a more complex understanding.
Well, it did give me a lot of information I had never heard of and then even though I don’t really know a lot about the French market and stuff. I feel like I could still apply my American consumer schemas to that new information. And so, I could kind of adapt my own perspective to the new facts that ChatGPT was giving me. (Student 14)
Comparing consulting websites and using ChatGPT, one student found that ChatGPT’s ability to analyze information and synthesize diverse viewpoints allowed them to reframe their interpretations and construct a more situated understanding rather than relying solely on conclusions they might reach on their own. Their reflection suggests an awareness of the limits of their own interpretive process and a tolerance for ambiguity. They recognized the challenge of trusting ChatGPT’s insights and of interpreting cultural preferences, acknowledging that their subjective nature makes verification difficult. This comment reflects an emerging capacity for dialectical thinking as they began to think between and across interpretive frameworks.
It’s made it much easier to think about problems from other perspectives because it takes away that whole step of having to make the analysis. If I were to just use a search engine, looking at French websites and then reading what they have to say, that’s more like just knowledge gathering. And then you come to a conclusion of a perspective by putting everything you’ve read together. But with ChatGPT, it just kind of streamlines that process and it’s already read those websites or it has access to people who have been there training the AI, thinking these same things. So, it does kind of take the perspective of other people and already applies it. There are certain things I do feel more of a need verifying, and it’s definitely the things I look up that are more factual, because that’s more easily verified. But if it’s something like, this is what some people prefer, I can’t exactly just search that up. Which is hard, because you have to trust it, which is difficult bridge to cross there. To what extent do you? (Student 5)
Contrasting their own intuitions with ChatGPT’s culturally contextualized responses introduced one student to ideas and perspectives they “would never think of.” Their comment underscores how ChatGPT provided tailored, specific feedback that went beyond what they could access through introspection or a generic web search. This interaction challenged their assumptions as they actively compared viewpoints. Their openness to choosing a name for their product based on the information provided suggests dialectical thinking, as they considered and weighed culturally situated perspectives alongside their own.
Me coming up with an idea that I think is good versus asking ChatGPT how French consumers would respond to it will definitely be like differing opinions because I think my product is good, but I don’t know about other people so, it’s definitely brought a new perspective to that. It definitely gives a detailed and explained opinion on anything you ask it in a way that I would never think of and feedback on potential market response that I wouldn’t get from anywhere else. So, it definitely gives more reasoning to the question I’m looking for. It gave me a list of 10 potential names and then gave the meaning behind the name in French and why it would be a good name, why it would appeal to consumers, and I just looked through them and thought of which one would be best. (Student 9)
5. Discussion
The findings suggest that the students in this study, similar to the participants in Hellmich et al. (Reference Hellmich, Vinall, Brandt, Chen and Sparks2024) and Xiao and Zhi (Reference Xiao and Zhi2023), were skeptical of ChatGPT and its outputs. Concerns about source transparency, awareness of embedded bias, and recognition of ChatGPT’s tendency to reproduce dominant cultural worldviews led students to question its credibility and cultural positioning. They did not perceive ChatGPT as a legitimate source of intercultural knowledge, doubting its authority to “speak” and its capacity to culturally frame its responses. As Lim and Darvin (Reference Lim and Darvin2026) note, limited transparency surrounding generative AI training data and algorithmic configurations poses challenges for promoting multiple perspectives and intercultural competence in the L2 classroom, because students may struggle to maintain the critical distance needed to identify potential bias and limitations in AI-generated content. In this study, students did not accept ChatGPT’s responses uncritically; instead, they drew on their own resources (e.g. their past experiences using the tool) and actively questioned its capacity to authentically generate culturally “other” perspectives, noting its tendency to offer generic consensus or tailor responses to perceived users’ expectations.
However, although such skepticism reflects critical thinking and learner agency, it also points to limitations in ChatGPT’s capacity to support the expansion of L2 learners’ interpretive frameworks. The findings of this study echo concerns raised by Liu et al. (Reference Liu, Lee and Zhao2025) regarding the integration of generative AI tools such as ChatGPT in L2 learning, given their tendency to reproduce dominant linguistic or cultural norms, thereby limiting the visibility of local realities. Lacking familiarity with frames of reference associated with the French market, students felt ill-equipped to detect inaccuracies and cultural stereotypes and expressed uncertainty about the cultural accuracy of ChatGPT’s responses. As a result, some students either dismissed the information or engaged with it superficially. Without encounters with unfamiliar or divergent perspectives, dissonance did not emerge, limiting the conditions necessary to question or (re)evaluate cultural assumptions and frames of reference. These findings suggest that a certain amount of trust in the source of information may be a crucial condition for generating the kind of dissonance that can decenter taken-for-granted assumptions by sustaining engagement with unfamiliar perspectives and supporting interpretive expansion.
The data also reveal that despite general skepticism toward ChatGPT, the interactive and conversational nature of the tool created specific conditions that supported students’ engagement in perspective shifting and cognitive flexibility. By positioning themselves subjectively within culturally unfamiliar roles, students were able to experiment with an emic stance and experiment with alternative viewpoints. For some, ChatGPT served as a springboard for dialectical thinking by presenting unexpected ideas that challenged their assumptions. As students recognized the limits of their own perspectives, their tolerance for ambiguity appeared to increase, and they began to think dialectically, moving between and across cultural frameworks rather than within a single perspective. Liu et al. (Reference Liu, Lee and Zhao2025) identify prompt literacy as a key dimension of learners’ AI practices. In this study, some students demonstrated emerging prompt literacy by crafting scenario-based prompts that elicited culturally situated responses. This enabled more active, co-creative engagements with ChatGPT, reflecting a form of shared agency in which students and ChatGPT collaboratively constructed meaning. Such dynamic partnerships allowed students to explore alternative perspectives and roles, suggesting that cognitive flexibility may be fostered through the dialogic and iterative nature of interactions with ChatGPT.
6. Conclusion
The goal of this study was to investigate whether interacting with ChatGPT offered a context that enabled students to expand their interpretive frameworks and shift between perspectives. Despite the small number of participants, the data analysis highlights the complexity of using ChatGPT to support the development of interculturality in the L2 curriculum and exposes a key pedagogical challenge: helping students see unfamiliarity not as a threat to validity but as a space for cultural interpretation and reflection, where ambiguity is embraced rather than resolved. The data point to the need for scaffolding the process. Structured activities such as roundtable discussions may support students in formulating prompts that elicit multiple perspectives, interrogating dissonances, and addressing doubts about AI-generated responses. ChatGPT also appears capable of disrupting initial certainties by introducing doubt, which may prompt students to move beyond their default assumptions and begin shifting between interpretive frames as they negotiate between their own assumptions and alternatives. This form of cognitive flexibility may represent a more realistic and achievable outcome than perspective transformation. This study suggests that ChatGPT-mediated interactions can support incremental shifts in interpretive positioning and cultivate tolerance for ambiguity.
The findings underline the importance of designing pedagogical activities that scaffold interactions in ways that foster subjective positioning and dialectical reflection. Prompt revision workshops, guided experimentation with scenario-based prompts, and reflective debriefing activities to examine how students evaluated the credibility and cultural positioning of ChatGPT responses may help students exercise interpretive control, critically evaluate AI-generated content, and balance trust and skepticism. Attending to cultural specificity, for example, by comparing ChatGPT responses with texts produced within the target culture, can further support nuanced interpretation. The study points to directions for future research. Studies could explore how ChatGPT might be used to generate dissonance around culturally familiar concepts or issues, or how different types of instructional scaffolding influence engagement with AI-generated perspectives and the development of interpretive flexibility. In the context of intercultural teaching and learning, ChatGPT’s pedagogical promise may reside less in what it knows and more in how it helps students think differently about what they think they know, serving not as a source of cultural knowledge but as a tool to cultivate the ability to shift between interpretive frames.
Data availability statement
The data that support the findings of this study are available within the article.
Acknowledgements
I would like to thank my research assistant, Warren Mark Matthews, for his valuable assistance in conducting and transcribing student interviews.
Authorship contribution statement
Isabelle Drewelow: Conceptualization, Methodology, Funding acquisition, Project administration, Resources, Supervision, Data curation, Formal analysis, Writing – original draft, Writing – review & editing.
Funding disclosure statement
This research was supported by a University of Alabama Faculty Teaching Fellowship.
Competing interest statement
The author declares no competing interests.
Ethical statement
The study received ethical approval from the institutional review board of the University of Alabama. Participation was voluntary, and all participants provided written informed consent prior to data collection. Confidentiality and anonymity of participants were maintained throughout the research.
GenAI use disclosure statement
The research assistant used Google Pinpoint to transcribe the audio files of each interview into text (March and April 2025). The author used Grammarly to check grammar, spelling, and sentence clarity during the manuscript preparation process (August 2025).
About the author
Isabelle Drewelow is an associate professor of French and applied linguistics at the University of Alabama. Her research explores the critical interplay between intercultural understanding, experiential and transformative pedagogical design, and the role of positive emotions in fostering learner engagement and personal growth.
Open access
