Designing an AI Roleplay System for Contextualized Career Guidance in Mexico

This paper presents the design of a ChatGPT-based roleplay system for career guidance among upper secondary students in Mexico, motivated by university dropout linked to mismatched expectations and contextual constraints such as commuting distance, costs, and family finances. Its primary aim is to offer a reproducible methodological account of prompt engineering rather than to claim vocational effectiveness. The system was refined through seven prompt iterations and is organized around a four-phase conversational architecture, an eight-dimension student profile, and a five-stage narrative simulation that closes with a data-sharing step.

The formative evaluation involved eight matched responses drawn from 13 initial participants, all remote upper secondary students in the State of Mexico. Its results are descriptive and non-inferential; an apparent rise in career certainty from 50% to 62.5% is explicitly cautioned against being read as an intervention effect. The most consistent finding was a usability barrier in the ChatGPT conversation-export step, which likely disadvantaged the least digitally prepared participants. The authors also acknowledge as a significant limitation the absence of formal ethics approval and of written or parental consent.

The work's scope is geographically confined to Mexico and to a small, single-region sample, and it addresses a specific segment of upper secondary education. Its contribution lies chiefly in documenting design decisions and surfacing practical barriers, such as digital literacy and data collection, that any future deployment of conversational AI in career guidance would need to address.