Context-based Reflection Support Counseling Agent
T. Shinozaki, Yukiko Yamamoto, K. Takada, Setsuo Tsuruta · 2012
The success rate of IT system development projects has remained as low as 30 %, resulting in that more than 60 % of workers have suffered from problems. This paper proposes a software agent to replace rudimentary-level human counselors. The agent uses a counseling knowledge whose domain is limited to IT workers or students. Counseling differs from learning support whose problems are clear and well-understood by teachers. Furthermore, teachers fundamentally know their solutions. It also differs from psychotherapy whose problems as well as solutions are less clear and much more general (i.e., not domain-specific). Utilizing domain knowledge focused on IT related counseling, our proposed counseling agent can: 1) support identification of the client's problem by building up mutual trust and empathy, and 2) promote the client's further clarification of their problems towards self-awareness of solutions by means of reflection solicited by ELIZA-like but context-dependent dialogue. Exploiting the context as well as the limited domain of problems makes the counseling software agent easily realized without acquiring and managing a tremendous amount of knowledge. The preliminary experimental results for six students showed that the agent can replace rudimentary-level human counselors. Namely, though incorporated with just limited knowledge, it can promote clients to enable more self-reflection e.g. by continuing 3 times longer dialogue than ELIZA and more self-awareness e.g. according to questionnaires, sufficient in the proposed agent but insufficient in ELIZA.