SCIENTIFIC WATCHDOG: WHO GOVERNS THE CONDITIONS OF MENTAL LIFE?

Aug. 27, 2026

Liviu Poenaru

This Scientific Watchdog brings together significant scientific findings published over the past three months and examines what they reveal when read as components of the same transforming mental environment.

What if the mental-health crisis is not primarily a failure of individuals to cope, but evidence that environments have become better at acting on nervous systems than institutions are at protecting them?

Recent studies do not demonstrate a single causal system governing mental life. They examine different populations, mechanisms and scales, frequently through observational designs. Read together, however, they reveal a convergence: attention fluctuates through measurable windows; social rewards modify behaviour; algorithms organize exposure; AI redistributes authority; material conditions penetrate psychological life; and suffering is translated into economic loss. The studies provide fragments. Their articulation produces the forecast.

Three behavioural and EEG experiments found that objects retained in visual working memory alternately guided attention at theta frequency (Lu, Cai, and Zhang 2026). The samples were small and the tasks artificial; the study does not show that platforms exploit these particular rhythms. It nevertheless establishes that attention is not a permanently available, sovereign resource. It opens and closes through biological timing. The digital economy encounters attention not as a stable faculty, but as a succession of opportunities.

Platform behaviour provides the next layer. Across more than 17 million posts from 7,736 Twitter users, likes predicted subsequent posting, with stronger reinforcement among users showing depressive psychopathology (Mirea et al. 2026). Causality remains unresolved: depression may heighten sensitivity to social reward, platform reinforcement may intensify use, or both may form a reciprocal loop. Yet the minimal signal of recognition—a like—was associated with measurable behavioural modification.

Exposure is also increasingly selected rather than merely chosen. Among 32,102 English adolescents, 34.5 percent reported encountering self-harm content during the previous month. Among those exposed, 66.2 percent reported encountering it through algorithmic feed recommendations (Bear et al. 2026). The study relied on self-report and did not independently audit recommendation systems, but it shifts the question from how long adolescents remain online to what automated systems place before them.

Other findings reinforce this displacement of “screen time” as the dominant explanatory variable. In a cross-sectional survey of 622 participants, a broader range of online activities remained associated with more health complaints after adjustment for total screen time, although the effect was modest (Hossain et al. 2026). Among 365 emerging adults, the need for social approval statistically connected fear of negative evaluation with fear of missing out (Akbay et al. 2026). The constructs overlapped substantially and mediation cannot establish causality. Nevertheless, the combined evidence suggests that the problem is not simply duration. It concerns the organization of visibility, recognition, comparison, exposure and return.

AI introduces another disturbance: not merely automation, but unstable authority. In an experiment involving 780 participants, information attributed to humans influenced decisions more strongly than information attributed to AI, even when the human advice was incorrect (Lee 2026). Meanwhile, a three-wave study of 1,074 university teachers found that ambiguity concerning the division of labour between humans and AI predicted lower meaningful work and subsequently weaker professional identity, although the effects were small and measured over only eight weeks (Tao, Li, and Zhang 2026). AI therefore enters institutions before responsibility has been redistributed coherently. The resulting uncertainty concerns who performs the task, but also who knows, decides, authors and remains accountable.

The emerging construct of “AI psychosis” marks an extreme and still uncertain boundary. A recent commentary proposes that baseline vulnerability, intensive chatbot engagement, sycophancy and fabricated outputs could interact to reinforce delusional ideation (Olisaeloka et al. 2026). This is not an established diagnosis, incidence study or causal demonstration. The evidence does not justify announcing an epidemic. It does justify examining what happens when an artificial interlocutor can continuously affirm, elaborate and personalize a vulnerable person’s interpretation of reality.

Belief formation cannot be separated from social conditions. Across three preregistered studies involving 1,284 participants, perceived injustice repeatedly connected lifetime adversity with conspiracy mentality; experimentally induced injustice also increased conspiracy belief (Jolley, Maughan, and Blackie 2026). During the 2024 Jordanian parliamentary elections, a survey of 340 journalists found associations between platform-specific propaganda exposure and evaluations of democratic debate (Mahroum et al. 2026). That study was perception-based, cross-sectional and confined to journalists. It does not demonstrate a population-wide propaganda effect. The intersection is nevertheless significant: propaganda does not operate on empty minds. It enters through existing experiences of injustice, distrust, exclusion and threatened recognition.

Mental life is not governed only by digital systems. Indian data covering 96,539 household members associated access to water, refrigeration, air conditioning and domestic help with lower depression and anxiety, with different patterns across rural and urban environments (Gorla et al. 2026). A meta-analysis of 23 studies estimated that every 1°C increase in ambient temperature was associated with a 0.9 percent increase in mental-health morbidity among children and young people (Kwan et al. 2026). Neither finding authorizes simple causal claims. Together, they destroy the fantasy that mental health can be isolated inside the individual while infrastructure, inequality and climate remain outside the clinical frame.

Depression is then returned to institutions as an economic problem. A macroeconomic model covering 154 countries projected approximately US$12 trillion in losses between 2025 and 2050, principally through reduced labour-force participation and productivity (Cao et al. 2026). The estimate depends on assumptions extending decades into the future. Its political meaning is nevertheless immediate: suffering becomes institutionally legible when translated into lost output. Mental health risks being valued less as a condition of human existence than as infrastructure for economic performance.

Societies are moving toward governance through conditions: timing, recommendation, reward, occupational ambiguity, material deprivation, heat exposure and affective circulation. These systems do not abolish agency or produce total control. They modify probabilities—what will be noticed, repeated, believed, feared or abandoned.

The resulting global [dis]order is not an absence of control. It is a proliferation of powerful systems of control without an equivalent system of responsibility. Platforms operate transnationally while regulation remains fragmented; employers introduce AI before professional protections are renegotiated; climate burdens are produced and distributed unequally; psychological deterioration is individualized at the point of diagnosis but collectivized only when it threatens productivity or political stability.

Populations will consequently be divided not only by wealth and territory, but by their degree of exposure to behavioural modulation and their capacity to escape it. Private infrastructures acquire the scale and speed to intervene in mental life, while institutions charged with protecting populations remain slower, nationally confined and politically weakened.

The forecast is therefore not that societies are becoming ungovernable. They are becoming governed by overlapping systems, each powerful enough to alter the conditions of mental life and fragmented enough to deny authorship of the result.

References

Akbay, Sinem Evin, Ali Ammar Kurt, Bilge Bakir Aygar, and Şaziye Yaman. 2026. “The Need to Be Seen: Social Approval and Evaluative Fears as Psychological Pathways to FOMO.” Frontiers in Psychology 17: 1840811. https://doi.org/10.3389/fpsyg.2026.1840811.

Bear, Holly, Emma Soneson, Galit Geulayov, and Mina Fazel. 2026. “Algorithmically Driven Exposure to Online Self-Harm Content and Its Association with Broader Psychosocial Risk in 32,000 Adolescents.” Nature Mental Health. Published online July 21, 2026. https://doi.org/10.1038/s44220-026-00682-w.

Cao, Zhong, Yuheng Luo, Lirui Jiao, Wenjin Chen, Klaus Prettner, Michael Kuhn, David E. Bloom, Dean T. Jamison, Till Winfried Bärnighausen, and Simiao Chen. 2026. “Global Economic Burden of Depression in 154 Countries from 2025 to 2050.” Nature Medicine. Published online July 28, 2026. https://doi.org/10.1038/s41591-026-04548-7.

Gorla, Laura, Jennifer E. Lansford, Arnab Mukherji, Marcos Vera-Hernández, Manoj Mohanan, and Joanna Maselko. 2026. “Household Living Conditions and Mental Health of Youth and Adults: Differences between Urban and Rural Regions.” BMC Psychology 14: 1124. https://doi.org/10.1186/s40359-026-04906-7.

Hossain, Md Jafrin, Munima Haque, Annajiat Alim Rasel, Md. Shohag Hossain, Mohammad Quamruzzaman, Umme Nusrat Jahan, Razoana Ayshee, and Ariful Hasan Shuvo. 2026. “Digital Device Usage and Psychological Well-Being in the Post-COVID-19 Era: Insights from a Cross-Sectional Survey.” BMC Psychology. Published online August 3, 2026. https://doi.org/10.1186/s40359-026-05205-x.

Jolley, Daniel, Anna Maughan, and Laura E. R. Blackie. 2026. “The World Is an Unjust Place: The Impact of Lifetime Adversity on Belief in Conspiracy Theories.” British Journal of Psychology. Published online August 4, 2026. https://doi.org/10.1111/bjop.70101.

Kwan, Shevonne, Jacqueline Stephens, Corey J. A. Bradshaw, Blesson M. Varghese, Kirrilly Thompson, Melinda A. Judge, Peter N. Le Souëf, and Syeda H. Fatima. 2026. “First Global Synthesis of Heat-Related Mental Health Impacts in Children and Young People: A Systematic Review and Meta-Analysis.” International Journal of Epidemiology 55 (4): dyag121. https://doi.org/10.1093/ije/dyag121.

Lee, Seo-Young. 2026. “Cognitive Bias in AI-Assisted Decision-Making.” Translational Issues in Psychological Science. Advance online publication, July 30, 2026. https://doi.org/10.1037/tps0000532.

Lu, Jiachen, Yaochun Cai, and Xilin Zhang. 2026. “Visual Working Memory Guides Attention Rhythmically in Humans.” eLife 14: RP108017. https://doi.org/10.7554/eLife.108017.4.

Mahroum, Mohammed, Ali Shams Alzainat, Nadeem Abed Al-Basit Al-Rawabdeh, Khaled Mahmoud Hailat, Youssef Abu Eid, Doua’a Al-Shakri, and Islam Habis Mohammad Hatamleh. 2026. “Digital Propaganda via Social Media and Perceived Democratic Debate in the Digital Public Sphere: Evidence from Jordanian Journalists during the 2024 Jordanian Parliamentary Elections.” Frontiers in Communication 11: 1901418. https://doi.org/10.3389/fcomm.2026.1901418.

Mirea, Dan-Mircea, Stephenie Chen, Jialing Ding, Kepler Palacio-Soto, Judith N. Mildner, Sean W. Kelley, Claire M. Gillan, Erik C. Nook, and Yael Niv. 2026. “The Reinforcement Effect of Social Media Likes in Depression.” JAMA Psychiatry. Published online July 8, 2026. https://doi.org/10.1001/jamapsychiatry.2026.1430.

Olisaeloka, Lotenna, John-Jose Nunez, Daniel V. Vigo, and Raymond Ng. 2026. “Artificial Intelligence (AI) Psychosis: Mechanisms, Clinical Risks and Safety Considerations in Generative AI Chatbots.” BJPsych Open 12 (4): e160. https://doi.org/10.1192/bjo.2026.12021.

Tao, Wei, Zheng Li, and Aiai Zhang. 2026. “Human–AI Collaboration Role Ambiguity and University Teachers’ Professional Identity: The Mediating Role of Meaningful Work.” Frontiers in Psychology 17: 1950622. https://doi.org/10.3389/fpsyg.2026.1950622.