Background & Statement of aims:
Generative AI presents new opportunities for discipline-specific, on-demand learning support, particularly in complex domains such as chemistry and physiology. This presentation reports on a cross institutional Monash-Warwick collaborative project that develops and evaluates custom GPTs designed to scaffold student problem solving in pharmacy and pharmaceutical science. Using the Goldilocks Help framework [1] as a pedagogical foundation, we constructed course-specific GPTs that offer Socratic prompts to support metacognitive engagement, while minimising the risk of reliance and over dependence on AI-generated answers.

Methods:
The implementation process involved iterative refinement and rigorous testing, both for accuracy and for Socratic capability. Our mixed-methods study investigates how students interact with these custom GPTs compared to open-access chatbots, drawing on usage data, perception surveys, and think-aloud interviews. Preliminary findings suggest that students adopt distinct prompting strategies depending on their familiarity with the subject matter, problem-solving mindset (instrumental or concept-driven), and the perceived authority of the chatbot. The custom GPTs were more likely to elicit structured engagement aligned with learning objectives, while open tools tended to promote trial-and-error approaches.

Results:
We discuss how the design of prompt scaffolds and model behaviour can influence student cognition, affect, and regulation during problem solving. We also highlight how student co-design contributed to the tool’s usability and perceived relevance.

Conclusion: As a practical take-away for others looking to design customised chatbots for their courses, we offer a skeleton template for a system prompt and simple steps to customise it for various contexts. This work has broader implications for the development of GenAI-driven support tools in STEM education and offers a scalable model for embedding discipline-aligned pedagogical principles into generative AI applications.

References:

1. Yuriev E, Naidu S, Schembri LS, & Short JL. Scaffolding the development of problem-solving skills in chemistry: guiding novice students out of dead ends and false starts. Chem. Educ. Res. Pract. 2017; 18(3), 486-504. doi.org/10.1039/C7RP00009J

Clinical decision-making1 is an important competency for pharmacy students, particularly in managing complex conditions such as venous thromboembolism (VTE). Simulation platforms like MyDispense2 provide opportunities for students to practice clinical decision-making and be assessed on this skill in an environment whereby they can make mistakes without affecting patient safety. This project aimed to explore the relationship between student performance and their clinical decision-making behaviours during a MyDispense VTE assessment, regarding how productive and unproductive student actions affect their clinical performance. 15 third-year pharmacy students were selected at random from three performance categories: low (<50%), mid (50–80%), and high (>80%). Students completed a VTE MyDispense assessment involving history-taking, review of discharge summaries and pathology reports, and therapeutic decision-making. MyDispense activity logs were analysed and behaviours were classified as productive, neutral, or unproductive. Correlations between student behaviours and performance were analysed using linear regression and ANOVAs. No significant correlations were found between assessment scores and time taken (p=0.083), discharge summary access (p=0.334), or counselling (p=0.554). A significant positive correlation was found between assessment scores and productive behaviours (r=0.682, p=0.005), with high-performing students completing more productive behaviours than low-performing students (p=0.006). Unproductive behaviours were negatively correlated with performance (r=-0.682, p=0.005), with low-performing students having more unproductive behaviours (p=0.002). The study demonstrates that assessment results depend heavily on the number of productive and unproductive actions students perform which supports the need for clinical reasoning education to enhance their decision-making abilities. This suggests that simulation platforms such as MyDispense can be valuable tools for identifying and addressing gaps in clinical reasoning. Further qualitative research is underway. MyDispense cases of increasing difficulty³ were developed by pharmacists from community, hospital, and academia for each year level (1–4). Students will complete the cases in think-aloud interviews, with video and transcript data abductively analysed using the Clinical Decision-Making Model¹ and Walker et al.’s⁴ cognitive and metacognitive framework.

1. Mertens JF et al., Res Social Adm Pharm. 2024;20(2):105-14.
2. Khera HK et al. J Pharm Policy Pract. 2023;16(1):110.
3. Spark MJ et al., Curr Pharm Teach Learn. 2022;14(10):1246-55.
4. Walker S et al., Am J Pharm Educ. 2023;87(4):ajpe8817.

Background:
Therapeutic reasoning is a core competency for person-centred care and is best developed through supervised experiential learning. Inconsistent coaching, limited feedback, and supervisor variability limit effective supervision. A generative artificial intelligence (GenAI) therapeutic reasoning co-coaching tool has the potential to support learners; however, little is known about healthcare professionals’ acceptance of such a tool.

Aim
To gather perceptions from healthcare learners on the potential use of a GenAI tool for coaching therapeutic reasoning including key characteristics such a tool should possess.

Methods
Semi-structured interviews with healthcare learners were conducted between January and April 2025. Transcripts were coded using a semi-inductive approach guided by an a priori framework based on Cognitive Apprenticeship theory and the Unified Theory of Acceptance and Use of Technology (UTAUT) [1,2]. Two independent coding teams (n=3 in each team) analysed the transcripts. Human Research Ethics Committee approval was obtained.

Results
Seven interviews were conducted, including pharmacy interns (n=3), nurse practitioners (n=2), an independent prescribing pharmacist (n=1), and a medical practitioner (n=1). 364 unique codes were generated; high interrater reliability was achieved (Cohen’s κ = 0.859; 84.8% agreement). Participant experiences/opinions aligned strongly with the theoretical frameworks. Coaching features such as articulation, domain knowledge, and scaffolding were highly valued from the participants human coach and desired in the GenAI tool. GenAI was seen as a promising co-coach, especially for its accessibility and ability to prompt critical thinking, although concerns were raised about data privacy and relational limitations. Key features desired in a GenAI tool included adaptability, reference to trusted clinical sources, and a capacity to tailor responses to the learner’s stage and style. A blended model combining human and GenAI coaching was favoured.

Discussion
Findings suggest that healthcare professionals recognise the potential of GenAI tools to support therapeutic reasoning, provided these tools include characteristics expected of an effective supervisor.

References:
1. Collins, A., Brown, J. S., & Holum, A. (1991). Cognitive apprenticeship: Making thinking visible. American Educator, 15(3), 6–11.
2. Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User Acceptance of Information Technology: Toward a Unified View. MIS Quarterly, 27(3), 425–478. https://doi.org/10.2307/30036540

Background and Aims: Language barriers pose a significant challenge to equitable healthcare delivery, influencing patient understanding, adherence, satisfaction, and health outcomes [1, 2]. This is particularly relevant in Malaysia, a multiracial society where Bahasa Malaysia is the primary language of patient interaction, yet English dominates tertiary education [3]. Consequently, pharmacy students may have limited opportunities to practise culturally and linguistically appropriate communication, potentially compromising patient-centred care. This study aimed to develop and validate (i) a questionnaire to assess students’ perception and preparedness and (ii) a generative artificial intelligence (GenAI)-supported simulation tool to enhance Malay language competency and preparedness for diverse patient interactions among Bachelor of Pharmacy students at Monash University Malaysia.
Methods: Local contextualised cases incorporating hospital and community pharmacy were adopted and developed for the ATLAS (Authentic Teaching & Learning Application Simulations) platform. A student questionnaire was developed to assess perceived effectiveness of the simulations in improving communication skills and readiness for practice. Content validity was evaluated by three experts for relevance, importance, and clarity. Face validation (n=10) involved qualitative feedback, which informed item refinement prior to pilot testing.
Results: Four GenAI avatars were developed and iteratively refined through internal trial testing. The questionnaire demonstrated excellent content validity, with S-CVI/Ave values of 0.96 (relevance), 0.96 (importance), and 0.97 (clarity). Qualitative feedback from face validation supported the clarity, contextual relevance, and comprehensibility of the instrument.
Conclusion: This formative work supports the feasibility and validity of integrating GenAI simulations into pharmacy education to enhance culturally responsive communication.
References
1.Ng CB, Chang CT, Ong SY, Mahmud M, Lee LC, Chew WY, et al. Awareness, expectation and satisfaction towards ward pharmacy services among patients in medical wards: a multi-centre study in Perak, Malaysia. BMC Health Serv Res. 2021;21(1):1175.
2.Mosiur RM, Mehar SM. English and Malay language policy and planning in Malaysia. Training, Language and Culture. 2021;5(4):36-46.
3.Brottman MR, Char DM, Hattori RA, Heeb R, Taff SD. Toward cultural competency in health care: a scoping review of the diversity and inclusion education literature. Acad Med. 2020;95:803–13.