Kostadin Kushlev

Is a higher frequency of general AI use associated with greater anxiety? Do these associations vary by country and cultural context? Using a sample of 34,275 individuals distributed across 35 countries, we found that, on average, higher usage of AI was associated with higher anxiety. The size of the association was small (β = 0.035, p = .001). However, this association varied significantly across countries, from β = +0.288 in South Africa to β = -0.299 in China. National AI governance readiness and cultural orientation (individualism, power distance) did not explain this cross-national variation. The variation was, however, partially explained by a country's aggregate optimism about AI. In countries where more respondents viewed AI as doing more harm than good, heavier AI use was associated with substantially higher anxiety, whereas in more optimistic countries, heavier AI use was not related to anxiety. This provides preliminary evidence that societal discourse surrounding AI shapes how AI use relates to anxiety, with implications for public communication and policy efforts about AI, as well as for research seeking to understand the general AI–mental health link.

Concerns that AI homogenizes human thinking appear at odds with findings that LLM writing is perceived as more creative than human writing. We propose that LLMs enhance surface-level diversity while simultaneously homogenizing underlying ideas. Four large-scale natural experiments tested this preregistered “semantic disjunction” hypothesis in the high-stakes context of college admissions. Comparing personal statements written by 372,793 applicants after-versus-before ChatGPT’s 2022 release, multi-level semantic analyses revealed increasing surface diversity of words simultaneous with decreasing conceptual diversity of sentences (within essays) and whole documents (between essays). Three controlled experiments provided causal evidence that AI use produced semantic disjunction, including in applicants from the natural experiment, and indicated an apparent root mechanism: a positive association between surface and conceptual diversity in humans was weakened or reversed in LLM writing. Despite conceptual homogenization, raters perceived post-ChatGPT essays as more creative. Semantic disjunction explained this apparent paradox: increasing word-level diversity explained increasing creativity judgments, overriding effects of decreased conceptual diversity, even when raters (human and AI) considered many essays together. Avoiding over-reliance on word-level surface diversity by considering multiple semantic levels is likely to support more complete understanding of AI’s effects on human thinking. Disjunctive trends were widespread, but were strongest in minoritized applicants.

Research on social digital technologies like smartphones and social media has grown exponentially, yet consensus on their effects on well‐being remains elusive. I introduce the Costs and Opportunities of Technology in Context (COTeC) framework, focused on defining and integrating the processes by which social technologies impact well‐being indirectly via their opportunity costs. COTeC specifies at least three such mechanisms: displacement (i.e., time reallocation: time spent on devices leaves less time for other activities), substitution (i.e., qualitative replacement: technology offers alternatives to more challenging, but often more rewarding, activities), and interference (i.e., real‐time attention reallocation: social technologies tax cognitive resources, reallocate attention, and disrupt ingrained social processes and expectations). Though evidence suggests that these mechanisms primarily undermine digital well‐being, the COTeC framework posits that their effects depend on contextual affordances: on what is being displaced, substituted, or interfered with. I consider the practical implications of the framework for policy, practice (e.g., digital detox interventions), and emerging technologies like artificial intelligence (AI). In short, COTeC aims to organize existing evidence of opportunity‐cost processes (e.g., phubbing vs. technoference), build on existing content‐focused and person‐focused approaches, and accelerate discoveries with practical implications.

Can mental health apps be a tool in tackling the declining mental health in the United States and around the globe? Our review of the literature shows that mental health apps can improve symptoms of some disorders, such as depression and anxiety, when compared to no treatment. Such evidence of efficacy, however, is lacking in populations that are underserved by existing mental health services, including youth, seniors, men, and racial minorities. Additionally, a large number of apps claim to be evidence-based without any publicly available peer-reviewed evidence. This unregulated marketplace risks misleading consumers and exacerbating health disparities. We propose multi-level policy solutions, including oversight by the Food and Drug Administration (FDA), voluntary adoption of scientific disclosure standards by app stores, and the creation of accessible, AI-enhanced databases to help consumers identify validated tools. Mental health apps promise to expand access to care, but without stronger evidence and regulation, they may deepen—rather than bridge—the mental health treatment divide.

Generative AI systems, especially Large Language Models (LLMs) such as ChatGPT, have recently emerged as significant contributors to creative processes. While LLMs can produce creative content that might be as good as or even better than human-created content, their widespread use risks reducing creative diversity across groups of people. In the present research, we aimed to quantify this homogenizing effect of LLMs on creative diversity, not only at the individual level but also at the collective level. Across three preregistered studies, we analyzed 2,200 college admissions essays. Using a novel measure—the diversity growth rate—we showed that each additional human-written essay contributed more new ideas than did each additional GPT-4 essay. Notably, this difference became more pronounced as more essays were included in the analysis and persisted despite efforts to enhance AI-generated content through both prompt and parameter modifications. Overall, our findings suggest that, despite their potential to enhance individual creativity, the widespread use of LLMs could diminish the collective diversity of creative ideas.

Smartphone overuse has been shown to negatively impact well-being. Digital detox interventions have been proposed as a method of curbing smartphone overuse; however, evidence is scant on the effectiveness of sustainable digital detox habits. The present study explores whether short, daily 30-min digital detox practices are effective in improving well-being. In a sample of 640 college students, we randomly assigned half to a control condition and the other half to one of three digital detox conditions: no use after waking up, no use before going to bed, and no use at a time of their choosing. In a daily diary design, we assessed participants’ daily affect over 7 days while the manipulation lasted. Results indicated that participants found it difficult to comply even with such brief digital detox practices and that there was no difference in daily affect between those who were assigned an intervention compared to the control group. However, when students successfully practiced digital detox before bed, they used their phones less and felt better the following day compared to control participants, and practicing digital detox in the morning more often was associated with better daily affect. Future research should examine how to help people practice digital detox more effectively.