Dhavan V. Shah

Despite overwhelming scientific consensus, climate change is a politicized issue in American politics (Chinn et al., 2020; Egan et al., 2022; McCright & Dunlap, 2011). Partisanship predicts environ...

Background: Online communities such as Reddit can reveal critical information needs (CINs) among people who discuss opioid use for chronic or acute pain under conditions of stigma and uncertainty. Methods: We used a mixed-method design to analyze 3428 unique Reddit posts from 1 January to 31 December 2022. Sentence-BERT embeddings and K-means clustering identified eight dominant themes, followed by qualitative analysis of a random subsample of 400 posts. Results: The most prevalent theme was seeking alternatives to opioids for pain relief (28.00%), followed by medications for chronic pain (19.84%) and kratom use experiences (14.64%). Qualitative findings showed recurrent concerns about addiction, stigma in clinical encounters, withdrawal, treatment access, and peer-to-peer discussion of kratom, tianeptine, cannabis, and other substitute substances. Conclusions: The study suggests an unmet need for accessible, nonstigmatizing, and clinically grounded information on pain management, tapering, and lower-risk alternatives to opioid use. These results have implications for addiction prevention, harm reduction, and patient-facing communication strategies for clinical practice and community health.

This study investigates the efficacy of AI-generated visuals and potential credibility cues in correcting health-related misinformation. Using a pre-registered, factorial experiment ( N = 1,751), we tested two AI-generated visual enhancements (visual exemplars, infographics) and two credibility boosters (source tagging, partisan neutrality in posting history). Findings revealed small but significant advantages of visual exemplars, but not infographics, over text-only corrections in reducing misbeliefs, primarily through mitigating psychological reactance. However, credibility cues did not significantly enhance correction effectiveness. Overall, the results suggest that AI-generated visual exemplars offer incremental persuasive benefits beyond textual correction alone.

We investigate how partisan media during and after the 2020 US presidential (Study 1) and 2022 midterm (Study 2) elections influenced preferences for unyielding politicians who do not compromise with partisan opponents. Our findings suggest that partisan media use may undermine willingness to support politicians who engage in deliberative compromises with opponents. This effect is likely driven by the tendency of partisan media to diminish willingness to engage in political listening and instigate moral considerations of the party’s policy goals.

Despite podcasts' increasing prominence in the contemporary media ecosystem, they have attracted limited scholarly attention. This study examines audience characteristics of the most popular podcast in the United States, The Joe Rogan Experience (JRE), situating it within the broader cultural context of the "manosphere." Leveraging two national probability samples collected in October 2022 (N=2,907) and November 2023 (N=1,000), this research explores the socio-demographic profile, political attitudes, cultural values, and media consumption behaviors of Rogan's listeners. The findings indicate that Rogan's audience transcends traditional partisan alignments, instead cohering around shared cultural orientations characterized by epistemic skepticism, institutional distrust, conspiracism, and distinct attitudes toward gender and race. This study underscores the importance of examining deeper cultural values alongside traditional political variables, enhancing our understanding of how influential media figures like Rogan shape and reflect broader ideological currents in an increasingly fragmented information environment.

BACKGROUND The integration of Large Language-and-Vision Assistant (LLaVA) models with food and nutrition data enables multimodal meal analysis and contextual dietary guidance. Despite this potential, the reliability and practical usefulness of such systems for supporting everyday dietary decision-making remain underexplored. OBJECTIVE This study introduces Purrfessor, an innovative AI chatbot designed to provide personalized dietary guidance through interactive, multimodal engagement. The study aimed to evaluate its performance in ingredient recognition and recipe generation. METHODS The Purrfessor chatbot was trained using a combination of the FoodData Central database from the USDA, the Recipe2img dataset featuring food images and corresponding recipes, a curated human-annotated dataset derived from Recipe1M, and a customized Q&A dialogue dataset. The system operates under a session-based, multi-turn interaction paradigm, with memory retained only within an active session and no cross-session memory persistence. We implemented a two-phase evaluation framework combining AI-based performance assessment and human scoring. RESULTS Purrfessor achieved a high average cosine similarity of 0.90 in ingredient recognition with human-coded references. In GPT-4.1–based evaluation of recipe generation quality, Purrfessor outperformed the raw LLaVA model across all evaluated dimensions, with the largest improvements in completeness (7.44 vs 6.52), consistency (8.90 vs 7.81), and clarity (9.13 vs 8.39). Overall recipe quality improved from 7.66 to 8.35. Automatic metrics indicated strong ingredient coverage (0.78) and moderate step complexity (0.74), with lower coherence (0.62) and temperature/time specification (0.59), yielding an overall structured score of 0.68. Human evaluators rated Purrfessor’s Q&A accuracy highly: correctness (M=8.71, SD=1.15), relevance (M=9.99, SD=0.10), and clarity (M=9.33, SD=0.68). Error analysis indicated that 56% of responses contained minor hallucinations (i.e., inclusion of inferred secondary details or invisible garnishes). At the same time, core food identification and overall recipe logic remained accurate. CONCLUSIONS Findings highlight the role of anthropomorphic chatbot design and multimodal AI in supporting engaging dietary health conversations. This study offers an example of AI-driven, evidence-based dietary guidance and underscores the potential of health chatbots to nudge informed health decision-making. Insights contribute to the development of digital health interventions and personalized health communication strategies, with implications for the design of engaging, user-centered AI health assistants. CLINICALTRIAL

The integration of Large Language and Vision Assistant models with food and nutrition data enables multimodal meal analysis and contextual dietary guidance. Despite this potential, the reliability and practical usefulness of such systems for supporting everyday dietary decision-making remain underexplored. This study introduces Purrfessor, an innovative artificial intelligence (AI) chatbot designed to provide personalized dietary guidance through interactive, multimodal engagement. The study aimed to evaluate its performance in ingredient recognition and recipe generation. The Purrfessor chatbot was trained using a combination of the FoodData Central database from the US Department of Agriculture (USDA), the Recipe2img dataset featuring food images and corresponding recipes, a curated human-annotated dataset derived from Recipe1M, and a customized question-and-answer dialogue dataset. The system operates under a session-based, multiturn interaction paradigm, with memory retained only within an active session and no cross-session memory persistence. We implemented a 2-phase evaluation framework combining AI-based performance assessment and human scoring. Purrfessor achieved a high average cosine similarity of 0.90 in ingredient recognition with human-coded references. In GPT-4.1-based (OpenAI) evaluation of recipe generation quality, Purrfessor outperformed the raw Large Language and Vision Assistant model across all evaluated dimensions, with the largest improvements in completeness (7.44 vs 6.52), consistency (8.90 vs 7.81), and clarity (9.13 vs 8.39). Overall recipe quality improved from 7.66 to 8.35. Automatic metrics indicated strong ingredient coverage (0.78) and moderate step complexity (0.74), with lower coherence (0.62) and temperature and time specification (0.59), yielding an overall structured score of 0.68. Human evaluators rated Purrfessor's question-and-answer accuracy highly: correctness (mean 8.71, SD 1.15), relevance (mean 9.99, SD 0.10), and clarity (mean 9.33, SD 0.68). Error analysis indicated that 56% of responses contained minor hallucinations (ie, inclusion of inferred secondary details or invisible garnishes). At the same time, core food identification and overall recipe logic remained accurate. Findings highlight the role of anthropomorphic chatbot design and multimodal AI in supporting engaging dietary health conversations. This study offers an example of AI-driven, evidence-based dietary guidance and underscores the potential of health chatbots to nudge informed health decision-making. Insights contribute to the development of digital health interventions and personalized health communication strategies, with implications for the design of engaging, user-centered AI health assistants.

Background: Despite the increasing popularity of mobile health (mHealth) technologies, little is known about which types of mHealth system engagement might affect the maintenance of antiretroviral therapy among people with HIV and substance use disorders. Objective: This study aimed to use longitudinal and detailed system logs and weekly survey data to test a mediation model, where mHealth engagement indicators were treated as predictors, substance use and confidence in HIV management were treated as joint mediators, and antiretroviral therapy adherence was treated as the outcome. We further distinguished the initiation and intensity of system engagement by mode (expression vs reception) and by communication levels (intraindividual vs dyadic vs network). Methods: Tailored for people with HIV living with substance use disorders, the mHealth app was distributed among 208 participants aged >18 years from 2 US health clinics. Supervised by medical professionals, participants received weekly surveys through the app to report their health status and medication adherence data. System use was passively collected through the app, operationalized as transformed click-level data, aggregated weekly, and connected to survey responses with a 7-day lagged window. Using the weekly check-in record provided by participants as the unit of analysis (N=681), linear regression and structure equation models with cluster-robust SEs were used for analyses, controlling within-person autocorrelation and group-level error correlations. Racial groups were examined as moderators in the structure equation models. Results: We found that (1) intensity, not initiation, of system use; (2) dyadic message expression and reception; and (3) network expression positively predicted medication adherence through joint mediators (substance use and confidence in HIV management). However, intraindividual reception (ie, rereading saved entries for personal motivation) negatively predicts medication adherence through joint mediators. We also found Black participants have distinct usage patterns, suggesting the need to tailor mHealth interventions for this subgroup. Conclusions: These findings highlight the importance of considering the intensity of system engagement, rather than initiation alone, when designing mHealth interventions for people with HIV and tailoring these systems to Black communities.

Background: In the United States, over 60% of adults aged 65 years or older have multiple chronic health conditions, with consequences that include reduced quality of life, increasingly complex but less person-centered treatment, and higher health care costs. A previous trial of ElderTree, an eHealth intervention for older adults, found socioemotional benefits for those with high rates of primary care use. Objective: This study tested the effectiveness of an ElderTree intervention designed specifically for older patients with multiple chronic conditions to determine whether combining it with primary care improved socioemotional and physical outcomes. Methods: In a nonblinded randomized controlled trial, 346 participants recruited from primary care clinics were assigned 1:1 to the ElderTree intervention or an attention control and were followed for 12 months. All participants were aged 65 years or older and had electronic health record diagnoses of at least three of 11 chronic conditions. Primary outcomes were mental and physical quality of life, psychological well-being (feelings of competence, connectedness, meaningfulness, and optimism), and loneliness. Tested mediators of the effects of the study arm (ElderTree vs active control) on changes in primary outcomes over time were 6-month changes in health coping, motivation, feelings of relatedness, depression, and anxiety. Tested moderators were sex, scheduled health care use, and number of chronic conditions. Data sources were surveys at baseline and 6 and 12 months comprising validated scales, and continuously collected ElderTree usage. Results: At 12 months, 76.1% (134/176) of ElderTree participants were still using the intervention. There was a significant effect of ElderTree (vs control) on improvements over 12 months in mental quality of life (arm × timepoint interaction: b=0.76, 95% CI 0.14-1.37; P=.02; 12-month ∆d=0.15) but no such effect on the other primary outcomes of physical quality of life, psychological well-being, or loneliness. Sex moderated the effects of the study arm over time on mental quality of life (b=1.33, 95% CI 0.09-2.58; P=.04) and psychological well-being (b=1.13, 95% CI 0.13-2.12; P=.03), with stronger effects for women than men. The effect of the study arm on mental quality of life was mediated by 6-month improvements in relatedness (α=1.25, P=.04; b=0.31, P<.001). Analyses of secondary and exploratory outcomes showed minimal effects of ElderTree. Conclusions: Consistent with the previous iteration of ElderTree, the current iteration designed for older patients with multiple chronic conditions showed signs of improving socioemotional outcomes but no impact on physical outcomes. This may reflect the choice of chronic conditions for inclusion, which need not have impinged on patients’ physical quality of life. Two ongoing trials are testing more specific versions of ElderTree targeting older patients coping with (1) chronic pain and (2) greater debilitation owing to at least 5 chronic conditions. Trial Registration: ClinicalTrials.gov NCT03387735; https://clinicaltrials.gov/study/NCT03387735 International Registered Report Identifier (IRRID): RR2-10.2196/25175