We consider a recommender system that takes into account the interplay between recommendations, the evolution of user interests, and harmful content.We model the impact of recommendations on user behavior, particularly the tendency to consume harmful content.We seek recommendation policies that establish a tradeoff between maximizing click-through rate (CTR) and mitigating harm.We establish conditions under which the user profile dynamics have a stationary point, and propose algorithms for finding an optimal recommendation policy at stationarity.We experiment on a semi-synthetic movie recommendation setting initialized with real data and observe that our policies outperform baselines at simultaneously maximizing CTR and mitigating harm.

Social media platforms like Facebook and YouTube connect people with communities that reflect their own values and experiences. People discover new communities either organically or through algorithmic recommendations based on their interests and preferences. We study online journeys users take through these communities, focusing particularly on ones that may lead to problematic outcomes. In particular, we propose and explore the concept of gateways, namely, entities associated with a higher likelihood of subsequent engagement with problematic content. We show, via a real-world application on Facebook groups, that a simple definition of gateway entities can be leveraged to reduce exposure to problematic content by 1% without any adverse impact on user engagement metrics. Motivated by this finding, we propose several formal definitions of gateways, via both frequentist and survival analysis methods, and evaluate their efficacy in predicting user behavior through offline experiments. Frequentist, duration-insensitive methods predict future harmful engagements with an 0.64–0.83 AUC, while survival analysis methods improve this to 0.72–0.90 AUC.

Social media platforms are being utilized by individuals with mental illness for engaging in self-disclosure, finding support, or navigating treatment journeys. Individuals also increasingly bring their social media data to psychotherapy consultations. This emerging practice during psychotherapy can help us to better understand how patients appropriate social media technologies to develop and iterate patient narratives -- the stories of patients' own experiences that are vital in mental health treatment. In this paper, we seek to understand patients' perspectives regarding why and how they bring up their social media activities during psychotherapy consultations as well as related concerns. Through interviews with 18 mood disorder patients, we found that social media helps augment narratives around interpersonal conflicts, digital detox, and self-expression. We also found that discussion of social media activities shines a light on the power imbalance and privacy concerns regarding use of patient-generated health information. Based on the findings, we discuss that social media data are different from other types of patient-generated health data in terms of supporting patient narratives because of the social interactions and curation social media inherently engenders. We also discuss privacy concerns and trust between a patient and a therapist when patient narratives are supported by patients' social media data. Finally, we suggest design implications for social computing technologies that can foster patient narratives rooted in social media activities.

“Time spent on platform” is a widely used measure in many studies examining social media use and well-being, yet the current literature presents unresolved findings about the relationship between time on platform and well-being. In this paper, we consider the moderating effect of people’s mindsets about social media — whether they think a platform is good or bad for themselves and for society more generally. Combining survey responses from 29,284 participants in 15 countries with server-logged data of Facebook use, we found that when people thought that Facebook was good for them and for society, time spent on the platform was not significantly associated with well-being. Conversely, when they thought Facebook was bad, greater time spent was associated with lower well-being. On average, there was a small, negative correlation between time spent and well-being and the causal direction is not known. Beliefs had a stronger moderating relationship when time-spent measures were self-reported rather than coming from server logs. We discuss potential mechanisms for these results and implications for future research on well-being and social media use.

For people diagnosed with mental health conditions, psychiatric hospitalization is a major life transition, involving clinical treatment, crisis stabilization and loss of access of social networks and technology. The period after hospitalization involves not only management of the condition and clinical recovery but also re-establishing social connections and getting back to social and vocational roles for successful reintegration - a significant portion of which is mediated by social technology. However, little is known about how people get back to social lives after psychiatric hospitalization and the role social technology plays during the reintegration process. We address this gap through an interview study with 19 individuals who experienced psychiatric hospitalization in the recent past. Our findings shed light on how people's offline and online social lives are deeply intertwined with management of the mental health condition after hospitalization. We find that social technology supports reintegration journeys after hospitalization as well as presents certain obstacles. We discuss the role of social technology in significant life transitions such as reintegration and conclude with implications for social computing research, platform design and clinical care.

involving resuming social roles and responsibilities, overcoming stigma and self-maintenance of the condition. Both clinical recovery and social reintegration need to go hand-in-hand for the overall well-being of individuals. However, research exploring social media for mental health has considered narrower, disjoint conceptualizations of people with mental illness - either as a patient or as a support-seeker. In this paper, we combine medical records with social media data of 254 consented individuals who have experienced a psychiatric hospitalization to address this gap. Adopting a theory-driven, Gaussian Mixture modeling approach, we provide a taxonomy of six heterogeneous behavioral patterns characterizing peoples' mental health status transitions around hospitalizations. Then we present an empirically derived framework, based on feedback from clinical researchers, to understand peoples' trajectories around clinical recovery and social reintegration. Finally, to demonstrate the utility of this taxonomy and the empirical framework, we assess social media signals that are indicative of individuals' reintegration trajectories post-hospitalization. We discuss the implications of combining peoples' clinical and social experiences in mental health care and the opportunities this intersection presents to post-discharge support and technology-based interventions for mental health.

In contrast to existing, static audience controls that map poorly onto users' ideal audiences on social networking sites, dynamic audience selection (DAS) controls can make intelligent inferences to help users select their ideal audience given context and content. But does this potential utility outweigh its potential intrusiveness? We surveyed 250 participants to model users' ideal versus their chosen audiences with static controls and found a significant misalignment, suggesting that DAS might provide utility. We then designed a sensitizing prototype that allowed users to select audiences based on personal attributes, content, or context constraints. We evaluated DAS vis-a-vis Facebook's existing audience selection controls through a counterbalanced summative evaluation. We found that DAS's expressiveness, customizability, and scalability made participants feel more confident about the content they shared on Facebook. However, low transparency, distrust in algorithmic inferences, and the emergence of privacy-violating side channels made participants find the prototype unreliable or intrusive. We discuss factors that affected this trade-off between DAS's utility and intrusiveness and synthesize design implications for future audience selection tools.

Many studies examining social media use rely on self-report survey questions about how much time participants spend on social media platforms. Because they are challenging to answer accurately and susceptible to various biases, these self-reported measures are known to contain error -- although the specific contours of this error are not well understood. This paper compares data from ten self-reported Facebook use survey measures deployed in 15 countries (N = 49,934) against data from Facebook's server logs to describe factors associated with error in commonly used survey items from the literature. Self-reports were moderately correlated with actual Facebook use (r = 0.42 for the best-performing question), though participants significantly overestimated how much time they spent on Facebook and underestimated the number of times they visited. People who spent a lot of time on the platform were more likely to misreport their time, as were teens and younger adults, which is notable because of the high reliance on college-aged samples in many fields. We conclude with recommendations on the most accurate ways to collect time-spent data via surveys.

A growing body of research is combining social media data with machine learning to predict mental health states of individuals. An implication of this research lies in informing evidence-based diagnosis and treatment. However, obtaining clinically valid diagnostic information from sensitive patient populations is challenging. Consequently, researchers have operationalized characteristic online behaviors as "proxy diagnostic signals" for building these models. This paper posits a challenge in using these diagnostic signals, purported to support clinical decision-making. Focusing on three commonly used proxy diagnostic signals derived from social media, we find that predictive models built on these data, although offer strong internal validity, suffer from poor external validity when tested on mental health patients. A deeper dive reveals issues of population and sampling bias, as well as of uncertainty in construct validity inherent in these proxies. We discuss the methodological and clinical implications of these gaps and provide remedial guidelines for future research.

Self-disclosure of stigmatized conditions is known to yield therapeutic benefits. Social media sites are emerging as promising platforms enabling disclosure around a variety of stigmatized concerns, including mental illness. What kind of behavioral changes precede and follow such disclosures? Do the therapeutic benefits of "opening up" manifest in these changes? In this paper, we address these questions by focusing on disclosures of schizophrenia diagnoses made on Twitter. We adopt a clinically grounded quantitative approach to first identify temporal phases around disclosure during which symptoms of schizophrenia are likely to be significant. Then, to quantify behaviors before and after disclosures, we define linguistic measures drawing from literature on psycholinguistics and the socio-cognitive model of schizophrenia. Along with significant linguistic differences before and after disclosures, we find indications of therapeutic outcomes following disclosures, including improved readability and coherence in language, future orientation, lower self preoccupation, and reduced discussion of symptoms and stigma perceptions. We discuss the implications of social media as a new therapeutic tool in supporting disclosures of stigmatized conditions.

Nowadays, brick-and-mortar stores are finding it extremely difficult to retain their customers due to the ever increasing competition from the online stores. One of the key reasons for this is the lack of personalized shopping experience offered by the brick-and-mortar stores. This work considers the problem of persona based shopping recommendation for such stores to maximize the value for money of the shoppers. For this problem, it proposes a non-polynomial time-complexity optimal dynamic program and a polynomial time-complexity non-optimal heuristic, for making top-k recommendations by taking into account shopper persona and her time and budget constraints. In our empirical evaluations with a mix of real-world data and simulated data, the performance of the heuristic in terms of the persona based recommendations (quantified by similarity scores and items recommended) closely matched (differed by only 8% each with) that of the dynamic program and at the same time heuristic ran at least twice faster compared to the dynamic program.