The Use of Artificial Intelligence–Based Conversational Agents (Chatbots) for Weight Loss: Scoping Review and Practical Recommendations
Bibliographic record
Abstract
BACKGROUND: Overweight and obesity have now reached a state of a pandemic despite the clinical and commercial programs available. Artificial intelligence (AI) chatbots have a strong potential in optimizing such programs for weight loss. OBJECTIVE: This study aimed to review AI chatbot use cases for weight loss and to identify the essential components for prolonging user engagement. METHODS: A scoping review was conducted using the 5-stage framework by Arksey and O'Malley. Articles were searched across nine electronic databases (ACM Digital Library, CINAHL, Cochrane Central, Embase, IEEE Xplore, PsycINFO, PubMed, Scopus, and Web of Science) until July 9, 2021. Gray literature, reference lists, and Google Scholar were also searched. RESULTS: A total of 23 studies with 2231 participants were included and evaluated in this review. Most studies (8/23, 35%) focused on using AI chatbots to promote both a healthy diet and exercise, 13% (3/23) of the studies used AI chatbots solely for lifestyle data collection and obesity risk assessment whereas only 4% (1/23) of the studies focused on promoting a combination of a healthy diet, exercise, and stress management. In total, 48% (11/23) of the studies used only text-based AI chatbots, 52% (12/23) operationalized AI chatbots through smartphones, and 39% (9/23) integrated data collected through fitness wearables or Internet of Things appliances. The core functions of AI chatbots were to provide personalized recommendations (20/23, 87%), motivational messages (18/23, 78%), gamification (6/23, 26%), and emotional support (6/23, 26%). Study participants who experienced speech- and augmented reality-based chatbot interactions in addition to text-based chatbot interactions reported higher user engagement because of the convenience of hands-free interactions. Enabling conversations through multiple platforms (eg, SMS text messaging, Slack, Telegram, Signal, WhatsApp, or Facebook Messenger) and devices (eg, laptops, Google Home, and Amazon Alexa) was reported to increase user engagement. The human semblance of chatbots through verbal and nonverbal cues improved user engagement through interactivity and empathy. Other techniques used in text-based chatbots included personally and culturally appropriate colloquial tones and content; emojis that emulate human emotional expressions; positively framed words; citations of credible information sources; personification; validation; and the provision of real-time, fast, and reliable recommendations. Prevailing issues included privacy; accountability; user burden; and interoperability with other databases, third-party applications, social media platforms, devices, and appliances. CONCLUSIONS: AI chatbots should be designed to be human-like, personalized, contextualized, immersive, and enjoyable to enhance user experience, engagement, behavior change, and weight loss. These require the integration of health metrics (eg, based on self-reports and wearable trackers), personality and preferences (eg, based on goal achievements), circumstantial behaviors (eg, trigger-based overconsumption), and emotional states (eg, chatbot conversations and wearable stress detectors) to deliver personalized and effective recommendations for weight loss.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.038 | 0.121 |
| Meta-epidemiology (narrow) | 0.002 | 0.002 |
| Meta-epidemiology (broad) | 0.005 | 0.008 |
| Bibliometrics | 0.018 | 0.012 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.007 | 0.008 |
| Open science | 0.004 | 0.004 |
| Research integrity | 0.005 | 0.004 |
| Insufficient payload (model declined to judge) | 0.006 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".