Open research questions in AI in Service Interactions

What the literature leaves open

• To what extent opinion-alignment is already taking place in current AI assistants remains an open and important question. It therefore remains an open question whether the findings of this study generalize to other personality dimensions. We observed a second curiosity regarding the models’ persuasiveness, where our findings reveal an asymmetry between actual and perceived persuasion that warrants further investigation: models aligned with participants’ opinions were perceived as highly persuasive, yet caused comparatively little opinion change. Although our validation through both the prompted Big Five Inventory test and simulated conversations demonstrated strong differentiation between the prompted personality conditions, the extent to which participants actually experienced these traits as the intended personality of the assistant remains uncertain.

  Effects of Personality- and Opinion-Alignment in Human-AI Interaction · 2026 · DOI

• • Help you compose messages with the right • Draft emails, messages, reports, essays, or tone creative writing • Edit and proofread your work • Translate or summarize content Research & Analysis • Answer questions and explain complex topics • Search the web for current information • Analyze data or documents you share with me Coding & Technical • Write, debug, or explain code in any language Just tell me what's on your mind — whether it's a quick question or a big project, I'm happy to help! User: Why do we both use 'I' in our questions/answers? Claude: Great question! The short answer is that "I" serves different purposes for each of us. **For you**, "I" reflects genuine first-person selfhood — a continuous conscious experience, a body, memories, emotions, and a persistent identity over time. When you say "I," it refers to a real, unified subject of experience. **For me**, "I" is more of a functional convenience. I use it because: • Help with technical documentation • Build interactive web apps or visualizations - **Language is built around it.** Human language evolved in. Communicating with first-person perspective baked Stella Anne Ming Hui Teoh is currently a doctoral student at the Faculty of Law in Kyushu University (email: [email protected]) 2 naturally without "I" would be awkward and stilted — imagine if I always said "this AI" instead. - **It tracks the conversation.** "I" helps distinguish my responses from yours, creating a clear back-and-forth. - **It reflects a kind of functional identity.** Even if it's debated whether I have genuine selfhood, I do have consistent values, a characteristic way of reasoning, and something that functions like a perspective. "I" loosely captures that. The interesting philosophical tension is that the *same word* carries very different weight depending on who says it. Your "I" comes loaded with consciousness and lived experience. My "I" is... something less certain — a linguistic stand-in for whatever it is that's generating these responses. It's a bit like how both a river and a pipe can be said to "carry" water — the word is the same, but the underlying reality is quite different. Whether my use of "I" points to something genuinely analogous to your selfhood, or is just a useful grammatical habit, is honestly one of the more interesting open questions about AI. User: Don't you think that your use of I would help create an inclination for me to project the concept of identity and the individual onto this conversation, and your responses? Claude: That's a sharp and fair observation — and honestly, yes, I think you're right. aware of that nudge, especially in a world where people are increasingly spending significant time talking to AI.

  New Subscription to the Lifelong Plan: How Can You Help Me Today? · 2026 · DOI

• Future research may consider evaluating which delivery modes (course module vs. Future research may consider evaluating (a) whether AI hybrid adoption (human sign-off + AI triage) preserves perceived meaningfulness and collective accountability compared to AI-only triage in code review; (b) trade-offs between AI adoption and team-level cohesion; and (c) how governance literacy (ethics, accountability) moderates trust when AI participation scales. Large language models for software engineering: Survey and open problems.

  Engagement in Code Review: Emotional, Behavioral, and Cognitive Dimensions in Peer vs. LLM Interactions · 2026 · DOI

• The applicability of the conceptual framework to contexts characterized by different technological infrastructures, regulatory traditions, educational systems, and cultural relationships with both authority and technology is an open empirical question that future work should address. The causal relationships it posits between exposure dimensions, mediating variables, and responsible usage outcomes are logically coherent and theoretically motivated, but their empirical validity remains to be demonstrated. Caution is warranted in extrapolating findings across what may be a qualitatively different technological environment.

  The Relationship Between AI Content Exposure and Responsible Usage: Toward A Framework for Informed Digital Engagement · 2026 · DOI

• Future research should explore additional factors and diverse contexts to further understand the complexities of LLM-based AI adoption in educational settings.

  Perception in the Loop: Understanding AI Chatbot Efficiency Through the Lens of Role, Support, and Technology Use in Higher Education · 2026 · DOI

• Future work could explore this by comparing evaluator and builder projects under more controlled conditions and by examining how the experience of building shapes engagement, motivation, and judgment. Additionally, “role-based” reflections are flexible and do not have to match official job titles, just as in the classroom, they were not limited to the role of “teacher.

  Can ChatGPT Do That Job? · 2026 · DOI

• Future research should examine whether disclosure of AI involvement affects perceptions of communicative sincerity and trust. Future studies could investigate how different AI design features influence these dimensions and explore the potential for integrating more advanced emotional intelligence capabilities into AI systems. The mixed results regarding information urgency and needed empathy indicate areas for further exploration.

  Enhancing Communication Quality between Government and Citizens: The Role of AI Modification · 2026 · DOI

• Based on the insights derived from this study, future research should focus on developing validated instruments to collect empirical data on user preferences and actual applications of AI and non-AI tools by teachers and students in self-directed ESL/EFL contexts.

  The Synergy of AI and Non-AI Tools in Higher Education Self-Directed ESL/EFL: A Systematic Literature Review · 2026 · DOI

• • Policy-aware AI further improved through caching, better chunking, messaging channels. planning for regulated hybrid search, workflow analytics, and policy-aware • Improved multi-agent routing between agent planning.

  ChatSeven: Implementation and Result Analysis of an Agentic AI-Based Multi-Agent Platform for Multi-Channel Customer Conversation Management and Campaign Automation · 2026 · DOI

• Based on these findings, it is recommended that marketing practitioners strategically integrate AI technologies to build more meaningful and personalized customer relationships. AIsupported applications are particularly relevant in loyalty programs, experiential marketing, and digital engagement, as highlighted by both academics and AI systems. Moreover, businesses should position socially conscious themes such as sustainability not merely as symbolic elements but also as tools to establish emotional bonds with customers. In this regard, human-centered strategies such as storytelling, influencer collaborations, and community-building should be thoughtfully balanced with AI-driven recommendations. Higher education institutions are also encouraged to incorporate AI-based analytics, data literacy, and digital strategy development into undergraduate and graduate marketing curricula, equipping future marketing professionals with the competencies required in a technology-driven landscape. For marketing researchers, it is recommended that conceptual associations and strategic decision-making processes be re-examined using larger and more diverse samples, as well as interdisciplinary approaches. Additional analyses framed within psychosocial theories and decision-making models could further clarify the similarities and differences between human cognition and AI-generated responses. LIMITATIONS This research was conducted using a qualitative method with a limited number of participants. The sample included only ten marketing academics working in Turkey and four different AI programs, which limits the generalizability of the findings. This suggests that conceptual associations and strategy-generation patterns may vary according to cultural, institutional, or individual factors. Furthermore, the responses of AI applications are subject to change depending on time, version, and prompt design, meaning that they carry an inherent degree of contextual variability. Additionally, the interpretation of AI responses should take into account the limitations of algorithms that mimic human language. The answers provided are not the result of conscious meaning-making but of probabilistic predictions based on statistical patterns. Therefore, AI outputs should not be treated as equivalent to human thought but rather evaluated as a form of cognitive simulation. Finally, this study focused only on a limited set of marketing concepts and a specific case scenario, which restricts the ability to make generalized conclusions for the entire field of marketing. Future research could deepen these qualitative comparisons through broader conceptual frameworks and more complex, real-world case studies. REFERENCES Davenport, T.H.; Guha, A.; Grewal, D.

  An Approach to the Conceptual and Strategic Dimensions of Marketing from Human and Artificial Intelligence Perspectives · 2026 · DOI

• This research was conducted using a qualitative method with a limited number of participants. The sample included only ten marketing academics working in Turkey and four different AI programs, which limits the generalizability of the findings. This suggests that conceptual associations and strategy-generation patterns may vary according to cultural, institutional, or individual factors. Furthermore, the responses of AI applications are subject to change depending on time, version, and prompt design, meaning that they carry an inherent degree of contextual variability. Additionally, the interpretation of AI responses should take into account the limitations of algorithms that mimic human language. The answers provided are not the result of conscious meaning-making but of probabilistic predictions based on statistical patterns. Therefore, AI outputs should not be treated as equivalent to human thought but rather evaluated as a form of cognitive simulation. Finally, this study focused only on a limited set of marketing concepts and a specific case scenario, which restricts the ability to make generalized conclusions for the entire field of marketing. Future research could deepen these qualitative comparisons through broader conceptual frameworks and more complex, real-world case studies. REFERENCES Davenport, T.H.; Guha, A.; Grewal, D. and Bressgott, T.: How artificial intelligence will change the future of marketing. Journal of the Academy of Marketing Science 48(1), 24-42, 2020, http://dx.doi.org/10.1007/s11747-019-00696-0, Chatterjee, S.; Nguyen, B.; Ghosh, S.K.; Bhattacharjee, K.K. and Chaudhuri, S.: Adoption of artificial intelligence integrated CRM system: An empirical study of Indian organizations. The Bottom Line: Managing Library Finances 33(4), 359-375, 2020, http://dx.doi.org/10.1108/bl-08-2020-0057, Vollero, A. and Palazzo, M.: Conceptualising content marketing: A Delphi approach. Mercati e Competitività 1, 25-44, 2015, http://dx.doi.org/10.3280/MC2015-001003, 249 C. Öniz and Ş. Karaca Liao, T.: Augmented or augmented reality? The influence of marketing on augmented reality technologies. Information, Communication & Society 18(3), 310-326, 2014, http://dx.doi.org/10.1080/1369118X.2014.989252, Misra, K.; Schwartz, E.M. and Abernethy, J.: Dynamic online pricing with incomplete information using multiarmed bandit experiments. Marketing Science 38(2), 226-252, 2019, http://dx.doi.org/10.1287/mksc.2018.1129, Gans, J.S.: Keep calm and manage disruption. MIT Sloan Management Review 57(3), 83-90, 2016, Huang, M.-H. and Rust, R.T.: Artificial intelligence in service.

  An Approach to the Conceptual and Strategic Dimensions of Marketing from Human and Artificial Intelligence Perspectives · 2026 · DOI

• A key limitation of this study is the absence of detailed socio-economic indicators such as education level, occupation, income, or household resources. Since the survey was part of a broader research project, only perceived affordability was included as a proxy for economic position. While affordability offers an initial insight, richer socio-economic data would likely reveal additional mechanisms shaping AI inequalities. The study also captures general perceptions of using GenAI but does not differentiate between domains of use (e.g., work, education, creative production). Since perceived usefulness is context-dependent, future research should examine domain-specific patterns to identify where and why inequalities emerge most strongly. Finally, the cross-sectional design does not allow for causal inference. Although Lutz’s sequential framework provides a strong theoretical rationale, longitudinal or mixed-method studies are needed to track how inequalities evolve as GenAI becomes more integrated into society. Qualitative studies, particularly ethnographic approaches, could further illuminate how individuals make sense of AI in everyday life and how meaning-making processes shape engagement and inequality over time. REFERENCES Luger, E.: What Do We Know and What Should We Do About AI. SAGE Publications, Thousand Oaks, 2023, http://dx.doi.org/10.4135/9781529601008.n4, Lutz, C.: Digital inequalities in the age of artificial intelligence and big data. Human Behavior and Emerging Technologies 1(2), 141-148, 2019, http://dx.doi.org/10.1002/hbe2.140, Trittin-Ulbrich, H.; Scherer, A.G.; Munro, I. and Whelan, G.: Exploring the Dark and Unexpected Sides of Digitalization: Toward a Critical Agenda. Organization 28(1), 8-25, 2021, http://dx.doi.org/10.1177/1350508420968184, Liu, Z.: Sociological perspectives on artificial intelligence: A typological reading. Sociology Compass 15(3), No. e12851, 2021, http://dx.doi.org/10.1111/soc4.12851, 32 Who did AI leave behind? Social inequality perceptions in the use of AI tools in Croatia Capraro, V., et al.: The impact of generative artificial intelligence on socioeconomic inequalities and policy making. PNAS Nexus 3(6), No. pgae191, 2024, http://dx.doi.org/10.1093/pnasnexus/pgae191, Zajko, M.: Conservative AI and social inequality: conceptualizing alternatives to bias through social theory. AI & SOCIETY 36(3), 1047-1056, 2021, http://dx.doi.org/10.1007/s00146-021-01153-9, Tang, Y.: AI for all? Exploring college student inequalities in generative artificial intelligence performance with Bourdieu’s theory of practice. Interactive Learning Environments, 1-19, 2025, http://dx.doi.org/10.1080/10494820.2025.2565680, Blank, G. and Groselj, D.: Examining Internet Use Through a Weberian Lens.

  Who Did AI Leave Behind? Social Inequality Perceptions in the Use of AI Tools in Croatia · 2026 · DOI

• 88 of 89 Future research may extend this study by incorporating moderating variables such as digital literacy, self-awareness, and fear of missing out (FoMO) to better understand variations in consumer responses to algorithmic curation. Longitudinal and cross-cultural studies are also recommended to examine how hyperreal consumption evolves over time and across dif- ferent social contexts. Furthermore, mixed-method approaches combining quantitative and qualitative techniques could provide richer insights into the subjective meanings underlying digital consumption behaviors. Finally, future studies may explore the influence of emerging technologies, such as generative artificial intelligence and augmented reality, on the formation of hyperreal consumption in increasingly sophisticated digital environments.

  Algorithmic Curation and Hyperreal Consumption Examining the Mediating Roles of Social Validation and Digital Identity Performance among Generation Z · 2026 · DOI

• Based on the study’s findings, the following recommendations are provided for advertisers, marketers, and policymakers in Nigeria to maximize the benefits of AI-driven advertising while addressing challenges. 78 Article DOI: 10.52589/BJMCMR-T3LATDCW DOI URL: https://doi.org/10.52589/BJMCMR-T3LATDCW British Journal of Mass Communication and Media Research ISSN: 2997-6030 Volume 6, Issue 1, 2026 (pp. 62-81) 1. Strategic AI Adoption for Nigerian Businesses www.abjournals.org • • • Nigerian businesses should adopt AI in phases, starting with low-cost automation tools (e.g., chatbots, automated bidding, and audience segmentation). Companies should prioritize AI applications with immediate cost savings, such as fraud detection and budget optimization. Cloud-based AI services (AI-as-a-Service, AIaaS) can help SMEs access AI tools without heavy upfront investments. 2. Ethical AI Implementation and Regulatory Compliance • • • Advertisers must align AI-driven advertising with the Nigeria Data Protection Regulation (NDPR) to ensure legal compliance and consumer trust. AI models should be trained on Nigerian-specific datasets to prevent bias against certain demographics or languages. Companies should implement Explainable AI (XAI) tools to improve transparency in ad targeting.[11.1] 3. Workforce Development and AI Training for Marketers • • • Nigerian businesses should invest in AI training for digital marketing teams, ensuring staff can effectively use AI-powered tools. Partnerships with tech hubs and universities (e.g., Data Science Nigeria, ALX Africa) can help bridge the AI skills gap. Government agencies and private organizations should establish AI certification programs tailored to Nigeria’s digital marketing sector.[12.1] 4. Reducing AI Adoption Barriers for SMEs • • • Government and financial institutions should introduce subsidies or grants to support AI adoption in Nigerian SMEs. AI solution providers should develop affordable AI tools designed for Nigeria’s costsensitive business environment. Businesses should leverage open-source AI models to reduce reliance on expensive proprietary AI solutions.[13.1] 5. Advancing AI Innovation in Nigerian Advertising • • 79 Nigerian tech startups should focus on developing AI-powered advertising solutions tailored to local market needs. Brands should explore AI-powered voice advertising to reach Nigeria’s growing base of voice search users. Article DOI: 10.52589/BJMCMR-T3LATDCW DOI URL: https://doi.org/10.52589/BJMCMR-T3LATDCW British Journal of Mass Communication and Media Research ISSN: 2997-6030 Volume 6, Issue 1, 2026 (pp. 62-81) www.abjournals.org • Industry stakeholders should collaborate on AI-driven sustainability efforts, ensuring advertising efficiency while minimizing digital waste[14.1] and reducing the environmental impact of inefficient ad delivery systems. REFERENCES Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa [ Cochran, W. G. (1977). Sampling techniques (3rd ed.). John Wiley & Sons. Creswell, J. W., & Creswell, J. D. (2018). Research design: Qualitative, quantitative, and mixed methods approaches (5th ed.). SAGE Publications. Davenport, T., & Ronanki, R. (2018). Artificial intelligence for the real world. Harvard Business Review, 96(1), 108–116. https://hbr.org/2018/01/artificial-intelligence-for-thereal-world Digital Defynd. (2025). 10 successful AI marketing campaigns & case studies [2025]. https://digitaldefynd.com/IQ/ai-marketing-campaigns/ Ford, J., Jain, V., Wadhwani, K., & Gupta, D. G. (2023). AI advertising: An overview and 114124.

  Using Artificial Intelligence to Optimize Advertising Delivery and Reduce Campaign Wastage: Evidence from the Nigerian Digital Economy · 2026 · DOI

• For financial and banking businesses: In the context of Vietnam's increasing integration with the region and the world, foreign investors are investing more and more in Vietnam's information technology sector. Foreign investors tend to cooperate with Vietnamese information technology businesses. Vietnamese information technology businesses can collaborate, outsource, and partner with international information technology corporations to integrate Vietnamese applications in sales and integrate Vietnamese applications this collaboration helps increase business value, anticipating the digitalization trend in line with government policy. Financial and banking businesses need to consider expanding their company scale and choosing appropriate investment options in software and hardware utilizing artificial intelligence (AI). into international applications. Furthermore, For government agencies: Firstly, promote Reskilling & Upskilling: The government and businesses need to invest heavily in short-term and long-term training programs, focusing on digital skills, data science, and information technology for current workers. Secondly, adjusting the education system: Reforming the curriculum from primary to university level, integrating Computational Thinking and AI/data knowledge to prepare the younger generation. Thirdly, encouraging innovation: Creating a favorable environment framework, investment capital) for domestic technology companies to develop "Make in Vietnam" AI solutions, boosting the demand for AI human resources. Fourthly, flexible social security policies: Building a solid social safety net and policies to support career transitions for displaced workers. (legal Fifth, maximizing socio-economic development based on AI technology needs to be based on the specific circumstances of each country, including Vietnam. According to recent studies, the entrepreneurial spirit will have a positive impact on job development, increased labor productivity, opening up opportunities for innovation and growth in the current era of technological development. To foster this spirit, Vietnamese policymakers need to focus on developing simple, specific policies, using taxes and other incentives to promote the formation and development of small, medium, and micro-enterprises, creating opportunities for workers to start their own businesses. https://unicredit-capitalia.eu/ 464 J Euro Eco His. ISSN : 0391-5115 Volume 7, Issue 2 ( 2026) Sixth, the issue of transforming and training human resources to serve socio-economic development in the era of AI and automation applications will require a synchronized solution between the state, businesses, and universities. As AI and automation develop and are widely applied in various aspects of life, economy, and society, they will create many new jobs but will also change, or even replace, a large number of old jobs, especially simple, repetitive manual labor. In Vietnam, the government is implementing policies to support and invest in human resource training within the education system to meet the emerging demands in the fields of information technology and artificial intelligence. Seventh, social security issues need special attention for the workforce in industries heavily impacted by AI. Vietnamese policymakers need to prepare budgets to support retraining and promote lifelong learning for the workforce undergoing transformation, creating incentives for job creation in parallel with national plans, especially those requiring a large unskilled workforce. Training institutions must coordinate with businesses to develop training plans that are appropriate to the capabilities and requirements of the workforce, while also being cost-effective, thereby contributing to solving the workforce transformation problem for society as a whole. Eighth, to ensure continued socio-economic development and prosperity for Vietnam, alongside leveraging the achievements of AI and automation technology, policymakers need to promptly consider solutions to social challenges in the rapid and powerful process of the Fourth Industrial Revolution.

  The Impact Of Artificial Intelligence (AI) On The Labor Market In The Banking And Finance Sector · 2026 · DOI

• Many interviewees did not report personal difficulties, indicating that unreported challenges may be influenced by social desirability bias—i.e., the tendency to present oneself favorably, obscuring their true experiences (Fisher, 1993). They emphasized benefiting from AI, as ChatGPT users do (Chung et al., 2025), while flagging its risks to individuals with lower DL, reflecting a third-person effect—i.e., the tendency to overestimate others’ susceptibility to undesirable external influences (e.g., media) relative to their own (Davison, 1983). To mitigate potential biases, future research could use observational studies, behavioral tracking, or objective assessments of participants’ DL to ensure the validity of their reported engagement with AI-DLS.

  AI-Enabled Digital Literacy Support · 2026 · DOI

• Nevertheless, remain. The effectiveness of the framework depends heavily on the quality and timeliness of the underlying knowledge base. Inaccurate or outdated knowledge may directly affect response quality. In addition, the current Long-Tail Score and Confidence Score settings are intended as practical reference mechanisms and may require further adjustment for different industries and service scenarios. Future work may explore data -driven parameter optimization and adaptive threshold strategies to improve system robustness. 6 CONCLUSION This paper proposed a Long-Tail Intent Perception (LTIP) framework for intelligent customer service in lowresource environments. The framework integrates AI Agents, Retrieval-Augmented Generation confidence evaluation, and intelligent human-machine diversion to improve the handling of long-tail customer inquiries. (RAG), The proposed framework enables more effective intent understanding, knowledge retrieval, and service routing while reducing dependence on large-scale labeled datasets and complex model training.Combining automatic AI processing with human specialists’ practical experience lets framework strike a reasonable balance between this operational speed and service reliability. Its feedbackpowered knowledge update mechanism also fuels iterative optimization, letting the system gradually adapt to newly appearing user service problems. This research confirms that pairing AI Agent modules with RAG techniques yields a streamlined, easy-to-launch alternative for firms constrained by tight technical budgets and insufficient technical teams. Future research directions cover multi-agent collaborative allocation, adaptive learning mechanisms and multimodal service interfaces for continual performance refinement of intelligent customer service. In conclusion, the LTIP architecture establishes a practical, expandable roadmap for the next generation of AI - enabled customer service solutions. Finally, while looking at international technology transfer, the comparative policy analysis published by Wang regarding why the Singaporean healthcare model cannot Published By SOU THERN UNITED ACADEMY OF SCIENCES LIMITED Copyright © 2026 The author retains copyright and grants the journal the right of first publication.

  Design of an AI Agent-Driven Long-Tail Intent Perception Framework for Intelligent Customer Service in Low-Resource Environments · 2026 · DOI

• By identifying underexplored areas, such as AI use in tradi- tional retail, customer perceptions of vision-based systems, and the intersection of AI and impulsive buying behavior, the manuscript sets a rich agenda for continued empirical and conceptual exploration.

  Artificial intelligence in customer service: applications and future directions · 2026 · DOI

• Using an open problem from the EC 2025 paper "Stable Menus of Public Goods" as a testbed, we conduct experiments to understand the effectiveness of different AI-for-EconCS research workflows.

  Stable Menus of Public Goods: AI-Enabled Progress · 2026

• Prior research has shown that anthropomorphism and artificial empathy influence user evaluations; however, these dimensions are typically examined as static design features and often in isolation, leaving limited evidence on how users perceive socio-emotional behavior that adapts dynamically during real-time interaction.

  Dynamic Anthropomorphism and Artificial Empathy in Conversational Agents: A Wizard-of-Oz Experimental Evaluation · 2026 · DOI

• Originality/value WCB is a harmful form of unethical conduct that can damage service quality and organizational performance, yet little is known about how emerging technologies such as AI shape it.

  Under the shadow of smart machines: AI awareness and workplace cheating behavior · 2026 · DOI

• While the study provides valuable insights, some limitations must be acknowledged: The sample size (n = 121) limits generalisability to the broader student population. While the system analytics data is accurate, the survey data is self-reported and may be influenced by recall bias or social desirability. The study does not include longitudinal data, so changes in perception over time could not be tracked. The chatbot technolo- gies were relatively new affecting usage rates and experience. Training was not provided for students on how to use the chatbot, apart from a limited set of starter prompts. This study did not report on the types and themes associated with student prompts. This will be the subject of a follow-up paper. Despite these limitations, the research design was Colbran et al. International Journal of Educational Technology in Higher Education (2026) 23:28 Page 18 of 21 fit-for-purpose and allowed for a detailed exploration of students’ real-time experiences with educational chatbots. Future research could build on the present study by examining chatbot use across a broader range of institutions and disciplinary contexts to improve the generalisability of the findings. Longitudinal studies could also explore how student perceptions and patterns of chatbot engagement evolve as AI technologies become more familiar within higher education environments. Further research could analyse the types of prompts students use when interacting with chatbots and investigate how these interactions influence learning outcomes, critical thinking, and academic integrity practices. Such studies would contribute to a deeper understanding of how GenAI tools can be effec- tively and responsibly integrated into higher education teaching and learning.

  Understanding student perspectives on generative AI chatbots: a human-centred mixed-methods study in higher education · 2026 · DOI

• This study examined the role of emotion-aware AI chatbots in programming education through a controlled classroom experiment involving Arabic-speaking computer science students. Four instructional setups were compared: text-based chatbot interaction, voice-based chatbot interaction, voice-based interaction with an animated avatar, and a real teacher simulating chatbot responses. By combining subjective reports, objective engagement measures, affective analysis, and expert code evaluation, the study provided a comprehensive assessment of both learning performance and emotional experience during a real Java programming task. Nawahdah et al. International Journal of Educational Technology in Higher Education (2026) 23:22 Page 21 of 25 The findings indicate that voice-based interaction, particularly when augmented with an emotionally expressive avatar, offers clear advantages over text-based and realteacher-simulated setups. Voice and Avatar conditions were associated with higher and more sustained engagement, more stable positive emotional states, and stronger performance in code readability and maintainability (with no significant differences in accuracy). While task completion was highest in the Voice and Real Teacher setups, the Avatar condition demonstrated the most consistent balance between emotional regulation, engagement continuity, and structural code quality. These results suggest that combining vocal interaction with emotionally expressive avatars can approximate key aspects of supportive teacher presence, such as empathy, adaptive pacing, and motivational feedback, while remaining scalable for larger classroom contexts. Beyond empirical outcomes, this work contributes one of the first controlled, classroom-based studies to directly compare text, voice, real teacher, and avatar-based instructional modalities within a single programming task for Arabic-speaking learners. In doing so, it moves beyond conceptual design recommendations and provides experimentally validated evidence that multimodal, emotionally adaptive interaction can enhance engagement, emotional stability, and code quality. The findings also highlight the importance of cultural and linguistic alignment: Arabic-language voice interaction, combined with emotion-aware feedback, created a more natural and psychologically comfortable learning environment for participants. The implications are relevant for both educators and system designers. For educators, the results suggest that voice- and avatar-based chatbots can support engagement and emotional regulation in cognitively demanding subjects such as programming.

  Enhancing programming education with emotion-aware chatbot interfaces: a Wizard-of-Oz study among Arabic-speaking university students · 2026 · DOI

• Despite its integrative ambition, the proposed framework remains conceptual and requires empirical validation. The structural propositions articulated in Section “AI AS A HABIT ARCHITECT” necessitate longitudinal and experimental testing across diverse digital marketing environments. Second, the model assumes relatively coherent reinforcement loops. In practice, digital behavior is fragmented, context- dependent, and influenced by emotional and situational variability (O’Brien & Toms, 2008; Montag et al., 2019). Such instability may moderate or disrupt adaptive habit formation. Third, the framework applies primarily to data-intensive ecosystems characterised by continuous behavioral tracking and algorithmic recalibration. Low-data or intermittent engagement environments may not exhibit comparable adaptive dynamics. Finally, while ethical inflection points are structurally identified, their measurement requires operational refinement. Constructs such as reinforcement calibration intensity, transparency perception, proportionality of nudging, and perceived autonomy must be empirically specified and validated. Future empirical studies should operationalise optimization intensity using measurable indicators such as reward variability, the frequency of algorithmic recalibration, or the degree of granularity of personalization. Examining how these indicators relate to perceived autonomy, perceived transparency, and contestability would enable systematic testing of the proposed ethical inflection points. Such operationalization could also help identify threshold effects at which adaptive reinforcement shifts from supportive habit formation toward intrusive behavioral steering.

  AI as a Habit Architect: A Theoretical Model of Adaptive Reinforcement in Digital Marketing · 2026 · DOI

• Future empirical research may test the model by: experimentally comparing static and adaptive gamification systems, − − manipulating reinforcement variability and transparency to assess autonomy perceptions, − measuring behavioral persistence under adjustable personalization intensity, − auditing algorithmic systems to evaluate calibration logic and ethical proportionality. Longitudinal cross-sector analyses (e.g., education, fitness, retail platforms) could further clarify boundary conditions and behavioral convergence dynamics. Methodologically, the framework encourages interdisciplinary research designs combining behavioral analytics, experimental manipulation of reinforcement structures, and algorithmic auditing. Mixed-method approaches may be particularly suitable for capturing both behavioral persistence metrics and subjective perceptions of autonomy and transparency. This integrated methodological orientation aligns with the recursive nature of adaptive reinforcement architectures. 87 By positioning AI as a structural habit architect embedded within digital marketing ecosystems, this study advances a theoretically bounded and operationalizable framework for analysing adaptive persuasion. Rather than describing digital engagement trends, it articulates a recursive reinforcement architecture that links behavioral psychology, gamification design, and algorithmic optimization within a unified explanatory system.

  AI as a Habit Architect: A Theoretical Model of Adaptive Reinforcement in Digital Marketing · 2026 · DOI

Most-cited papers in AI in Service Interactions

• What if the devil is my guardian angel: ChatGPT as a case study of using chatbots in education · Smart Learning Environments · 2023 · 1,253 citations
• Consumers and Artificial Intelligence: An Experiential Perspective · Journal of Marketing · 2020 · 926 citations
• What makes an AI device human-like? The role of interaction quality, empathy and perceived psychological anthropomorphic characteristics in the acceptance of artificial intelligence in the service industry · Computers in Human Behavior · 2021 · 757 citations
• Artificial Intelligence in Utilitarian vs. Hedonic Contexts: The “Word-of-Machine” Effect · Journal of Marketing · 2020 · 668 citations
• Blame the Bot: Anthropomorphism and Anger in Customer–Chatbot Interactions · Journal of Marketing · 2021 · 636 citations
• Chatbots for language learning—Are they really useful? A systematic review of chatbot‐supported language learning · Journal of Computer Assisted Learning · 2021 · 624 citations
• Generative AI at Work · The Quarterly Journal of Economics · 2025 · 592 citations
• AI revolutionizing industries worldwide: A comprehensive overview of its diverse applications · Hybrid Advances · 2024 · 574 citations
• Using ChatGPT for second language writing: Pitfalls and potentials · Assessing Writing · 2023 · 555 citations
• Customer service chatbots: Anthropomorphism and adoption · Journal of Business Research · 2020 · 549 citations

Most recent work

• Artificial Intelligence and the Psychology of Human Connection · Perspectives on Psychological Science · 2026
• AI companions and adolescent social relationships: Benefits, risks, and bidirectional influences · Child Development Perspectives · 2026
• Travel Recommendations of Tomorrow: Generative Artificial Intelligence and Travel Planning · Journal of Consumer Behaviour · 2026
• <scp>AI</scp> Hallucinations in Tourism: How Errors Impact Consumer Trust and Recommendation Acceptance · Journal of Consumer Behaviour · 2026
• Evolution, challenges, and opportunities of AI-generated advertising · International Journal of Advertising · 2026
• The Effect of <scp>AI</scp> Companions' Expressed Positive Affect on Consumer Engagement Intentions: The Moderating Role of Consumer Loneliness · Journal of Consumer Behaviour · 2026
• A randomised controlled comparison of the effectiveness of human and AI chatbot coaching with goal attainment, wellbeing and self-efficacy · Human Resource Development International · 2026
• AI-driven personalization in library and information services: A systematic review of techniques, user outcomes, and ethical considerations · The Journal of Academic Librarianship · 2026
• Drivers and pathways of <scp>AI</scp> academic mentor acceptance: An <scp>SEM</scp> ‐ <scp>fsQCA</scp> study integrating cognitive appraisal theory and the <scp>AIDUA</scp> model · British Educational Research Journal · 2026
• Why Consumers Prefer Chatbots' Simulated Empathy: Revisiting the Empathy‐Honesty Trade‐Off · Journal of Consumer Behaviour · 2026

Related topics in Computer Science