PAL to Lend a Helping Hand: Towards Building an Emotion Adaptive Polite and Empathetic Counseling Conversational Agent

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 12254–12271 July 9-14, 2023 ©2023 Association for Computational Linguistics PAL to Lend a Helping Hand: Towards Building an Emotion Adaptive Polite and Empathetic Counseling Conversational Agent Kshitij Mishra ∗ and Priyanshu Priya ∗ and Asif Ekbal Department of Computer Science and Engineering Indian Institute of Technology Patna, Bihta, Bihar, India {kshitij_1921cs23,priyanshu_2021cs26,asif}@iitp.ac.in Abstract The World Health Organization (WHO) has significantly emphasized the need for mental health care. The social stigma associated with mental illness prevents individuals from ad- dressing their issues and getting assistance. In such a scenario, the relevance of online coun- seling has increased dramatically. The feel- ings and attitudes that a client and a counselor express towards each other result in a higher or lower counseling experience. A counselor should be friendly and gain clients’ trust to make them share their problems comfortably. Thus, it is essential for the counselor to ade- quately comprehend the client’s emotions and ensure client’s welfare, i.e. s/he should adapt and deal with the clients politely and empathet- ically to provide a pleasant, cordial and person- alized experience. Motivated by this, in this work, we attempt to build a novel Polite and empAthetic counseLing conversational agent PAL. To have client’s emotion-based polite and empathetic responses, two counseling datasets laying down the counseling support to sub- stance addicts and crime victims are anno- tated. These annotated datasets are used to build PAL in a reinforcement learning frame- work. A novel reward function is formulated to ensure correct politeness and empathy prefer- ences as per client’s emotions with naturalness and non-repetitiveness in responses. Thorough automatic and human evaluation showcases the usefulness and strength of the designed novel reward function. Our proposed system is scal- able and can be easily modified with different modules of preference models as per need1. 1 Introduction The global prevalence of mental health disorders is estimated to affect 20% of the world population (Holmes et al., 2018). The COVID-19 pandemic ∗Jointly first authors. 1Data and codes can be accessed at https://github.com/Mishrakshitij/PAL.git and AI-NLP- ML_PAL. Figure 1: Example demonstrating the variation in polite and empathetic response of the agent depending on the client’s emotion. The text highlighted in pink color shows the client’s emotion. has further jeopardized the mental well-being of in- dividuals (Paredes et al., 2021). Further, the limited availability of experts (Organization et al., 2021) and the overwhelming healthcare burden (Huffman et al., 2021) have created a substantial demand for counseling services. Counselors are faced with a diverse audience, including individuals struggling with substance addiction and crime victims, who often experience mental health challenges such as anxiety, depression, and post-traumatic stress (Eisen et al., 1992; Ross and Peselow, 2012; Kil- patrick et al., 1985; Kilpatrick and Acierno, 2003). Therefore, pertinent and timely counseling support is crucial. In light of this, it is imperative to explore the potential benefits that counseling could derive from the advancements in artificial intelligence (AI) and natural language processing …

Content truncated to 500 words