Linguistic Complexity Loss in Text-Based Therapy

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 4450–4459 June 6–11, 2021. ©2021 Association for Computational Linguistics 4450 Linguistic Complexity Loss in Text-Based Therapy Jason Wei ˚ Kelly Finn Emma Templeton Thalia Wheatley Soroush Vosoughi Dartmouth College, Hanover, NH Department of Computer Science Neukom Institute Department of Psychological and Brain Sciences {jason.20, kelly.r.finn, emma.m.templeton.gr, thalia.p.wheatley, soroush}@dartmouth.edu Abstract The complexity loss paradox, which posits that individuals suffering from disease exhibit surprisingly predictable behavioral dynamics, has been observed in a variety of both human and animal physiological systems. The recent advent of online text-based therapy presents a new opportunity to analyze the complexity loss paradox in a novel operationalization: lin- guistic complexity loss in text-based therapy conversations. In this paper, we analyze linguistic complex- ity correlates of mental health in the online therapy messages sent between therapists and 7,170 clients who provided 30,437 correspond- ing survey responses on their anxiety. We found that when clients reported more anxiety, they showed reduced lexical diversity as esti- mated by the moving average type-token ratio. Therapists, on the other hand, used language of higher reading difficulty, syntactic complex- ity, and age of acquisition when clients were more anxious. Finally, we found that clients, and to an even greater extent, therapists, exhib- ited consistent levels of many linguistic com- plexity measures. These results demonstrate how linguistic analysis of text-based commu- nication can be leveraged as a marker for anx- iety, an exciting prospect in a time of both in- creased online communication and increased mental health issues. 1 Introduction The complexity loss paradox (Goldberger, 1997) posits that individuals suffering from a wide range of illnesses tend to exhibit surprisingly periodic and predictable dynamics in their behavior, even though the diseases themselves are often called dis-orders. The paradox exists in patterns of behavior from diving in penguins (Cottin et al., 2014) to social interactions in chimpanzees (Alados and Huffman, ˚Now AI Resident at Google. Dataset Exploratory Confirmatory Messages 2.6 million 0.7 million Survey responses 24,287 6,150 Clients 5,736 1,434 Therapists 1,608 889 �Survey responses / client 4.23 4.29 �Client text (words) / survey 1259 1295 �Therapist text (words) / survey 796 804 Median survey score (0-21) 8 8 Median time between surveys 21 days 21 days Table 1: Descriptive statistics for Talkspace online ther- apy conversations dataset. � indicates mean. 2000). In humans, the paradox has been observed in physiological systems from the indistinguish- able tremors of Parkinsonian patients (Parker et al., 2018) to the cyclic oscillations of white blood cell counts in leukemia patients (Malhotra and Salam, 1991), but how the paradox manifests in one of our most important behavioral outputs—language— has not been well-studied. In what form could the complexity loss paradox manifest in language? A line of psycholinguistics research, starting from the 1970s, has shown that the words people use can reveal important aspects of their mental health (Pennebaker et al., 2003). For instance, vague and qualified speech can pre- dict depression …

Content truncated to 500 words