Background: At any given time, most smokers in a population are ambivalent with no motivation to quit. Motivational interviewing (MI) is an evidence-based technique that aims to elicit change in ambivalent smokers. MI practitioners are scarce and expensive, and smokers are difficult to reach. Smokers are potentially reachable through the web, and if an automated chatbot could emulate an MI conversation, it could form the basis of a low-cost and scalable intervention motivating smokers to quit. Objective: The primary goal of this study is to design, train, and test an automated MI-based chatbot capable of eliciting reflection in a conversation with cigarette smokers. This study describes the process of collecting training data to improve the chatbot’s ability to generate MI-oriented responses, particularly reflections and summary statements. The secondary goal of this study is to observe the effects on participants through voluntary feedback given after completing a conversation with the chatbot. Methods: An interdisciplinary collaboration between an MI expert and experts in computer engineering and natural language processing (NLP) co-designed the conversation and algorithms underlying the chatbot. A sample of 121 adult cigarette smokers in 11 successive groups were recruited from a web-based platform for a single-arm prospective iterative design study. The chatbot was designed to stimulate reflections on the pros and cons of smoking using MI’s running head start technique. Participants were also asked to confirm the chatbot’s classification of their free-form responses to measure the classification accuracy of the underlying NLP models. Each group provided responses that were used to train the chatbot for the next group. Results: A total of 6568 responses from 121 participants in 11 successive groups over 14 weeks were received. From these responses, we were able to isolate 21 unique reasons for and against smoking and the relative frequency of each. The gradual collection of responses as inputs and smoking reasons as labels over the 11 iterations improved the F1 score of the classification within the chatbot from 0.63 in the first group to 0.82 in the final group. The mean time spent by each participant interacting with the chatbot was 21.3 (SD 14.0) min (minimum 6.4 and maximum 89.2). We also found that 34.7% (42/121) of participants enjoyed the interaction with the chatbot, and 8.3% (10/121) of participants noted explicit smoking cessation benefits from the conversation in voluntary feedback that did not solicit this explicitly. Conclusions: Recruiting ambivalent smokers through the web is a viable method to train a chatbot to increase accuracy in reflection and summary statements, the building blocks of MI. A new set of 21 smoking reasons (both for and against) has been identified. Initial feedback from smokers on the experience shows promise toward using it in an intervention.

Background: It has become possible to use data from a patient’s mobile phone as an adjunct or alternative to the traditional self-report and interview methods of symptom assessment in psychiatry. Mobile data–based assessment is possible because of the large amounts of diverse information available from a modern mobile phone, including geolocation, screen activity, physical motion, and communication activity. This data may offer much more fine-grained insight into mental state than traditional methods, and so we are motivated to pursue research in this direction. However, passive data retrieval could be an unwelcome invasion of privacy, and some may not consent to such observation. It is therefore important to measure patients’ willingness to consent to such observation if this approach is to be considered for general use. Objective: The aim of this study was to measure the ownership rates of mobile phones within the patient population, measure the patient population’s willingness to have their mobile phone used as an experimental assessment tool for their mental health disorder, and, finally, to determine how likely patients would be to provide consent for each individual source of mobile phone–collectible data across the variety of potential data sources. Methods: New patients referred to a tertiary care mood and anxiety disorder clinic from August 2016 to October 2017 completed a survey designed to measure their mobile phone ownership, use, and willingness to install a mental health monitoring app and provide relevant data through the app. Results: Of the 82 respondents, 70 (85%) reported owning an internet-connected mobile phone. When asked about installing a hypothetical mobile phone app to assess their mental health disorder, 41% (33/80) responded with complete willingness to install with another 43% (34/80) indicating potential willingness to install such an app. Willingness to give permissions for specific types of data varied by data source, with respondents least willing to consent to audio recording and analysis (19% [15/80] willing respondents, 31% [25/80] potentially willing) and most willing to consent to observation of the mobile phone screen being on or off (46% [36/79] willing respondents and 23% [18/79] potentially willing). Conclusions: The patients surveyed had a high incidence of ownership of internet-connected mobile phones, which suggests some plausibility for the general approach of mental health state inference through mobile phone data. Patients were also relatively willing to consent to data collection from sources that were less personal but expressed less willingness for the most personal communication and location data.

Background: The motivational interviewing (MI) approach has been shown to help move ambivalent smokers toward the decision to quit smoking. There have been several attempts to broaden access to MI through text-based chatbots. These typically use scripted responses to client statements, but such nonspecific responses have been shown to reduce effectiveness. Recent advances in natural language processing provide a new way to create responses that are specific to a client’s statements, using a generative language model. Objective: This study aimed to design, evolve, and measure the effectiveness of a chatbot system that can guide ambivalent people who smoke toward the decision to quit smoking with MI-style generative reflections. Methods: Over time, 4 different MI chatbot versions were evolved, and each version was tested with a separate group of ambivalent smokers. A total of 349 smokers were recruited through a web-based recruitment platform. The first chatbot version only asked questions without reflections on the answers. The second version asked the questions and provided reflections with an initial version of the reflection generator. The third version used an improved reflection generator, and the fourth version added extended interaction on some of the questions. Participants’ readiness to quit was measured before the conversation and 1 week later using an 11-point scale that measured 3 attributes related to smoking cessation: readiness, confidence, and importance. The number of quit attempts made in the week before the conversation and the week after was surveyed; in addition, participants rated the perceived empathy of the chatbot. The main body of the conversation consists of 5 scripted questions, responses from participants, and (for 3 of the 4 versions) generated reflections. A pretrained transformer-based neural network was fine-tuned on examples of high-quality reflections to generate MI reflections. Results: The increase in average confidence using the nongenerative version was 1.0 (SD 2.0; P=.001), whereas for the 3 generative versions, the increases ranged from 1.2 to 1.3 (SD 2.0-2.3; P<.001). The extended conversation with improved generative reflections was the only version associated with a significant increase in average importance (0.7, SD 2.0; P<.001) and readiness (0.4, SD 1.7; P=.01). The enhanced reflection and extended conversations exhibited significantly better perceived empathy than the nongenerative conversation (P=.02 and P=.004, respectively). The number of quit attempts did not significantly change between the week before the conversation and the week after across all 4 conversations. Conclusions: The results suggest that generative reflections increase the impact of a conversation on readiness to quit smoking 1 week later, although a significant portion of the impact seen so far can be achieved by only asking questions without the reflections. These results support further evolution of the chatbot conversation and can serve as a basis for comparison against more advanced versions.

This paper describes a low-cost, robust, and accurate remote eye-tracking system that uses an industrial prototype smartphone with integrated infrared illumination and camera. Numerous studies have demonstrated the beneficial use of eye-tracking in domains such as neurological and neuropsychiatric testing, advertising evaluation, pilot training, and automotive safety. Remote eye-tracking on a smartphone could enable the significant growth in the deployment of applications in these domains. Our system uses a 3D gaze-estimation model that enables accurate point-of-gaze (PoG) estimation with free head and device motion. To accurately determine the input eye features (pupil center and corneal reflections), the system uses Convolutional Neural Networks (CNNs) together with a novel center-of-mass output layer. The use of CNNs improves the system’s robustness to the significant variability in the appearance of eye-images found in handheld eye trackers. The system was tested with 8 subjects with the device free to move in their hands and produced a gaze bias of 0.72°. Our hybrid approach that uses artificial illumination, a 3D gaze-estimation model, and a CNN feature extractor achieved an accuracy that is significantly (400%) better than current eye-tracking systems on smartphones that use natural illumination and machine-learning techniques to estimate the PoG.

Background: The lack of access to mental health care could be addressed, in part, through the development of automated screening technologies for detecting the most common mental health disorders without the direct involvement of clinicians. Objective smartphone-collected data may contain sufficient information about individuals’ behaviors to infer their mental states and therefore screen for anxiety disorders and depression. Objective: The objective of this study is to compare how a single set of recognized and novel features, extracted from smartphone-collected data, can be used for predicting generalized anxiety disorder (GAD), social anxiety disorder (SAD), and depression. Methods: An Android app was designed, together with a centralized server system, to collect periodic measurements of objective smartphone data. The types of data included samples of ambient audio, GPS location, screen state, and light sensor data. Subjects were recruited into a 2-week observational study in which the app was run on their personal smartphones. The subjects also completed self-report severity measures of SAD, GAD, and depression. The participants were 112 Canadian adults from a nonclinical population. High-level features were extracted from the data of 84 participants, and predictive models of SAD, GAD, and depression were built and evaluated. Results: Models of SAD and depression achieved a significantly greater screening accuracy than uninformative models (area under the receiver operating characteristic means of 0.64, SD 0.13 and 0.72, SD 0.12, respectively), whereas models of GAD failed to be predictive. Investigation of the model coefficients revealed key features that were predictive of SAD and depression. Conclusions: We demonstrate the ability of a common set of features to act as predictors in the models of both SAD and depression. This suggests that the types of behaviors that can be inferred from smartphone-collected data are broad indicators of mental health, which can be used to study, assess, and track psychopathology simultaneously across multiple disorders and diagnostic boundaries.

Background: Objective and continuous severity measures of anxiety and depression are highly valuable and would have many applications in psychiatry and psychology. A collective source of data for objective measures are the sensors in a person’s smartphone, and a particularly rich source is the microphone that can be used to sample the audio environment. This may give broad insight into activity, sleep, and social interaction, which may be associated with quality of life and severity of anxiety and depression. Objective: This study aimed to explore the properties of passively recorded environmental audio from a subject’s smartphone to find potential correlates of symptom severity of social anxiety disorder, generalized anxiety disorder, depression, and general impairment. Methods: An Android app was designed, together with a centralized server system, to collect periodic measurements of the volume of sounds in the environment and to detect the presence or absence of English-speaking voices. Subjects were recruited into a 2-week observational study during which the app was run on their personal smartphone to collect audio data. Subjects also completed self-report severity measures of social anxiety disorder, generalized anxiety disorder, depression, and functional impairment. Participants were 112 Canadian adults from a nonclinical population. High-level features were extracted from the environmental audio of 84 participants with sufficient data, and correlations were measured between the 4 audio features and the 4 self-report measures. Results: The regularity in daily patterns of activity and inactivity inferred from the environmental audio volume was correlated with the severity of depression (r=−0.37; P<.001). A measure of sleep disturbance inferred from the environmental audio volume was also correlated with the severity of depression (r=0.23; P=.03). A proxy measure of social interaction based on the detection of speaking voices in the environmental audio was correlated with depression (r=−0.37; P<.001) and functional impairment (r=−0.29; P=.01). None of the 4 environmental audio-based features tested showed significant correlations with the measures of generalized anxiety or social anxiety. Conclusions: In this study group, the environmental audio was shown to contain signals that were associated with the severity of depression and functional impairment. Associations with the severity of social anxiety disorder and generalized anxiety disorder were much weaker in comparison and not statistically significant at the 5% significance level. This work also confirmed previous work showing that the presence of voices is associated with depression. Furthermore, this study suggests that sparsely sampled audio volume could provide potentially relevant insight into subjects’ mental health.