Artificially intelligent chatbots in digital mental health interventions: a review

In the citing paper

“Relative to chatbots that do not express empathy, chatbots that express empathy lead to more favorable ratings of companionship, i.e., the activities done together are perceived as more enjoyable or exciting (Boucher et al. 2021; Leite et al. 2013), and better mood after experiencing social exclusion (De Gennaro, Krumhuber, and Lucas 2020).”

In the citing paper

“However, the study of the applications of AI in mental healthcare remains in its infancy (6, 20).”

In the citing paper

“AI technology can help address these issues by supporting a versatile, scalable, and powerful approach that can analyze and respond to large amounts of data and adapt to individual healthcare needs (6, 17–23).”

In the citing paper

“Nevertheless, recent studies have shown promising results of AI applications including suicide risk assessment and risk prediction, identification of mental illness predictors, mental health monitoring, psychoeducation and psychotherapy delivery, therapist training, mental healthcare personalization, mental health triage and decision-making, and promoting treatment engagement (6, 17–23).”

In the citing paper

“However, because of the novelty of this technology, there are several unanswered questions and considerations such as limitations in the interpretation of language, biases in the interaction with patients from diverse backgrounds, and unanswered ethical, patient safety, and health policy considerations (6, 17–23).”

In the citing paper

“Moreover, despite the accessibility of online mental healthcare interventions, reduced therapist support or guidance in some online programs, have resulted in participants having low levels of engagement with their treatment and a high risk of dropping out (6).”

In the citing paper

“Additionally, the studies reporting on the use of ML algorithms found significant success in outcome prediction and feature detection to support online therapy delivery (6, 19, 28, 29, 31, 33, 83, 85, 86, 92–94).”

In the citing paper

“These reviews recognized the novelty of this technology and the scarcity of evidence but noted several potential benefits in terms of providing a low-resource intensity alternative for individuals experiencing less severe mental health symptoms (6, 17).”

In the citing paper

“For instance, two recent reviews by Boucher et al. (2021) and Abd-alrazaq et al. (2019), reported that chatbots had shown promising applications including care assistance, symptom screening and monitoring, and therapy delivery.”

In the citing paper

“This presents the interesting possibility of a potential therapeutic relationship between user and chatbot, which is something that self-guided eCBT has not been able to accomplish (6, 17–23, 74).”

In the citing paper

“The available literature supports the results obtained by this mixed-methods review, showing that AI-augmented interventions such as AI therapy chatbots can be promising, effective and accessible solutions for mental healthcare delivery (6, 17–23).”

In the citing paper

“As such, the relevant studies highlighted the importance of analyzing the large datasets produced by online mental healthcare interventions and the efficiency and accuracy of AI algorithms which in some cases were comparable to or better than human assessors (6, 19, 28, 29, 31, 33, 83, 85, 86, 92–94).”

In the citing paper

“Additionally, the scarcity of studies and low quality of evidence, including the lack of a control or adequate placebo group impacts the strength and reliability of the reported outcomes (6, 17, 18).”

In the citing paper

“Moreover, AI chatbots and language models have been inefficient at interpreting and responding to idioms, colloquialisms, slang, and metaphors.”

In the citing paper

“In a global survey of psychiatrists, around 83% of respondents felt that AI would not be able to match the level of empathetic care provided by a professional (6, 98).”

In the citing paper

“Moreover as mentioned by other reviews in the field, several of the available studies lack an appropriate comparator or control group which can directly impact the observed effect of the AI-augmented interventions and tools reported in the literature (6, 17–23, 52–54, 57, 58, 67, 72).”

In the citing paper

“At the same time, a growing body of research shows that interactions with chatbots can contribute to improving their users’ mental health and quality of life, especially if the specific needs of the user match with chatbot capabilities (22, 24, 27, 31, 34, 40, 49–56).”