Helping the Helper: Supporting Peer Counselors via AI-Empowered Practice and Feedback

TL;DR

This paper presents CARE, an AI-augmented training tool designed to empower online peer counselors by diagnosing context-specific MI strategies and generating exemplar responses in real time. Built on BERT for strategy prediction and DialoGPT for response generation, CARE augments training through a three-step pipeline (strategy prediction, strategy-conditioned response generation, safety filtering) and is evaluated via quantitative log analyses and qualitative user studies on 7 Cups. Results show CARE increases counselor engagement, yields longer and potentially more calibrated responses, and is perceived as straightforward and helpful, particularly for novices, while highlighting risks around disruption, nuance gaps, and content accuracy. The work argues for an augmentation-based design with human oversight, discusses ethical/societal considerations, and outlines a roadmap for broader deployment and long-term impact assessment across platforms and topics.

Abstract

Millions of users come to online peer counseling platforms to seek support. However, studies show that online peer support groups are not always as effective as expected, largely due to users' negative experiences with unhelpful counselors. Peer counselors are key to the success of online peer counseling platforms, but most often do not receive appropriate training.Hence, we introduce CARE: an AI-based tool to empower and train peer counselors through practice and feedback. Concretely, CARE helps diagnose which counseling strategies are needed in a given situation and suggests example responses to counselors during their practice sessions. Building upon the Motivational Interviewing framework, CARE utilizes large-scale counseling conversation data with text generation techniques to enable these functionalities. We demonstrate the efficacy of CARE by performing quantitative evaluations and qualitative user studies through simulated chats and semi-structured interviews, finding that CARE especially helps novice counselors in challenging situations. The code is available at https://github.com/SALT-NLP/CARE

Figures (14)

• Figure 1: Interface of how CARE empowers supporters by diagnosing which skill to use and suggesting responses.
• Figure 2: CARE system architecture. CARE consists of 8 next-strategy predictors, one next-response generator, and one inappropriate response classifier. It predicts the probability distribution of the next counseling strategies and generates suggested responses for each confident strategy independently. In the end, the list of strategies and responses is filtered and ordered decreasingly by the confidence of the predictors.
• Figure 3: The peer counselors' frontend view of the tool. We design CARE as a Suggestion Panel. The other components (Chat Information, Chat Window, and Text Input Box) mimic the frontend view of 7 Cups. CARE is comprised of three elements: 1) The blue words: Suggested MI counseling strategy to be used in the next peer counselor response. Users can hover over the strategy to see its 1-sentence description. 2) Black sentences in boxes: A generated next peer counselor response that utilized the MI strategy above it. Users can click the response to auto-fill the input text box for modification before sending. 3) Hide/Show example button: Toggle for collapsing and expanding Suggestion Panel.
• Figure 4: The responses to perception questions. a) A bar chart of the participants' answers to Likert scale questions regarding their perception of the suggested counseling strategies (strategies), example responses (examples), and future usage of CARE. b) A bar chart of participants' overall perception of CARE. The dashed lines mark the medians of responses.
• Figure 5: The responses to multi-select checkbox questions regarding preferences. The dashed lines mark the medians of responses.