We present Meena, a multi-turn open-domain chatbot trained end-to-end on data mined and filtered from public domain social media conversations. This 2.6B parameter neural network is simply trained to minimize perplexity of the next token. We also propose a human evaluation metric called Sensibleness and Specificity Average (SSA), which captures key elements of a human-like multi-turn conversation. Our experiments show strong correlation between perplexity and SSA. The fact that the best perplexity end-to-end trained Meena scores high on SSA (72% on multi-turn evaluation) suggests that a human-level SSA of 86% is potentially within reach if we can better optimize perplexity. Additionally, the full version of Meena (with a filtering mechanism and tuned decoding) scores 79% SSA, 23% higher in absolute SSA than the existing chatbots we evaluated.

Towards a Human-like Open-Domain Chatbot

{CI}, �3.59 to �1.40], �2.63 [95% CI, �3.73 to �1.54], and �1.63 [95% CI, �2.73 to �0.53] at 6, 12, and 18 months, respectively). At 12 months, response and remission rates (CALM vs UC) were 63.66% (95% CI, 58.95%-68.37%) vs 44.68% (95% CI, 39.76%-49.59%), and 51.49% (95% CI, 46.60%-56.38%) vs 33.28% (95% CI, 28.62%-37.93%), with a number needed to treat of 5.27 (95% CI, 4.18-7.13) for response and 5.50 (95% CI, 4.32-7.55) for remission. Conclusion For patients with anxiety disorders treated in primary care clinics, CALM compared with UC resulted in greater improvement in anxiety symptoms, depression symptoms, functional disability, and quality of care during 18 months of follow-up.

/pdf/delivery-of-evidence-based-treatment-for-multiple-anxiety-3hnhosvdvy.pdf

Delivery of evidence-based treatment for multiple anxiety disorders in primary care: a randomized controlled trial.

There is a growing interest in chatbots, which are machine agents serving as natural language user interfaces for data and service providers. However, no studies have empirically investigated people’s motivations for using chatbots. In this study, an online questionnaire asked chatbot users (N = 146, aged 16–55 years) from the US to report their reasons for using chatbots. The study identifies key motivational factors driving chatbot use. The most frequently reported motivational factor is “productivity”; chatbots help users to obtain timely and efficient assistance or information. Chatbot users also reported motivations pertaining to entertainment, social and relational factors, and curiosity about what they view as a novel phenomenon. The findings are discussed in terms of the uses and gratifications theory, and they provide insight into why people choose to interact with automated agents online. The findings can help developers facilitate better human–chatbot interaction experiences in the future. Possible design guidelines are suggested, reflecting different chatbot user motivations.

Why People Use Chatbots

Objective: To implement and test the effectiveness of a combined pharmacotherapy and cognitive-behavioral intervention for panic disorder tailored to the primary care setting. Design: Randomized, controlled study comparing intervention to treatment as usual. Setting: Six primary care clinics associated with 3 university medical schools, serving an ethnically and socioeconomically diverse patient population. Participants: Two hundred thirty-two primary care patients meeting DSM-IV criteria for panic disorder. Comorbid mental and physical disorders were permitted, provided these did not contraindicate the treatment to be provided and were not acutely life threatening. Intervention: Patients were randomized to receive either treatment as usual or an intervention consisting of a combination of up to 6 sessions (across 12 weeks) of cognitivebehavioral therapy (CBT) modified for the primary care setting, with up to 6 follow-up telephone contacts during the next 9 months, and algorithm-based pharmacotherapy provided by the primary care physician with guidance from a psychiatrist. Behavioral health specialists, the majority inexperienced in CBT for panic disorder, were trained to deliver the CBT and coordinated overall care, including pharmacotherapy. Main Outcomes Measures: Proportion of subjects remitted (no panic attacks in the past month, minimal anticipatory anxiety, and agoraphobia subscale score 10 on Fear Questionnaire) and responding (Anxiety Sensitivity Index score 20) and change over time in World Health Organization Disability Scale and short form 12 scores. Results: The combined cognitive-behavioral and pharmacotherapeutic intervention resulted in sustained and gradually increasing improvement relative to treatment as usual, with significantly higher rates at all points of both the proportion of subjects remitted (3 months, 20% vs 12%; 12 months, 29% vs 16%) and responding (3 months, 46% vs 27%; 12 months, 63% vs 38%) and significantly greater improvements in World Health Organization Disability Scale (all points) and short form 12 mental health functioning (3 and 6 months) scores. These effects were obtained in spite of similar rates of delivery of guideline-concordant pharmacotherapy to the 2 groups. Conclusion: Delivery of evidence-based CBT and medication using the collaborative care model and a CBTnaive, midlevel behavioral health specialist is feasible and significantly more effective than usual care for primary care panic disorder. Arch Gen Psychiatry. 2005;62:290-298

https://jamanetwork.com/journals/jamapsychiatry/articlepdf/208417/yoa40168.pdf

A Randomized Effectiveness Trial of Cognitive-Behavioral Therapy and Medication for Primary Care Panic Disorder

In the shades of the uncanny valley: An experimental study of human–chatbot interaction

Real conversations with artificial intelligence

The authors report on development, reliability, and findings of the Psychosocial Treatments Interview (PTI) to assess treatments reported by patients in a naturalistic study of the longitudinal course of anxiety disorders. The PTI ascertains frequency of different types of psychosocial treatments, based on patients' reports. The PTI showed good internal consistency and very good interrater reliability. At first 6-month follow-up, the most common modalities were supportive, medication discussion, and dynamic intervention. Combinations were common. Delivery of treatments differed by site. Overall, the PTI fills a methodological need for the assessment of the treatments reported by patients in naturalistic follow-up studies.

Ingrid G. Farreras

Papers

Real conversations with artificial intelligence

The psychosocial treatments interview for anxiety disorders. A method for assessing psychotherapeutic procedures in anxiety disorders.

The Historical Context for National Institute of Mental Health Support of American Psychological Association Training and Accreditation Efforts.

LIWC-ing at a Century of Introductory College Textbooks: Have the Sentiments Changed?

The Multidimensional Orientation Toward Dying and Death Inventory (MODDI-F): Factorial Validity and Reliability in a U.S. Sample