Implementing Cognitive Behavioral Therapy in Chatbots to Reduce Students’ Exam Stress using ChatGPT

Abstract

Undergraduate students frequently encounter academic stress, highlighting the need for accessible and effective coping mechanisms. This study develops an AI chatbot grounded in Cognitive Behavioral Therapy (CBT) principles to meet the mental health needs of these students. The primary goal is to devise and assess a chatbot that delivers personalized support for managing academic stress. The chatbot engages students in meaningful conversations, identifies cognitive distortions in their thought processes, and provides customized coping strategies. It leverages OpenAI’s GPT-3.5 turbo LLM for generating lifelike interactions and uses a vector database to store domain-specific knowledge. Responses are preprocessed with OpenAI embeddings for intent classification. Support Vector Machines (SVM), Multinomial Logistic Regression, and Random Forest models are used for classifying cognitive distortions, with SVM achieving the highest accuracy of $85.94 \%$. Intent classification is refined using cosine similarity, enhancing the chatbot’s conversational accuracy. Although SVM outperforms other models, ongoing challenges call for further research to refine the chatbot’s functionality and improve its impact on student well-being.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 3 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.