Dark loops: contagion effects, consistency and chemosocial matrices in psychedelic-assisted therapy trials

The paper situates itself in the recent revival of clinical research into psychedelic-assisted therapy (PAT), noting a rapid increase in trials and a concurrent rise in public hype and social movements around psychedelics. Earlier research has established promising clinical signals for indications such as depression, PTSD, anxiety in terminal illness and substance use disorders, but the authors argue that growing public enthusiasm, media coverage and participant networks risk reshaping the interventions and the conditions under which trial results are produced. Noorani and colleagues set out to formalise how these social dynamics interact with the logic of causal inference used in randomised controlled trials (RCTs). They introduce the anthropological concept of chemosociality—social bonds emerging from shared chemical exposure—and propose the term "dark loops" for feedback structures that form between trials, participants, communities and media, which are largely unmeasured in existing analyses. The paper focuses on PAT trials that combine psychedelic administration (MDMA and classical hallucinogens) with psychotherapy, explains how these phenomena can breach key causal assumptions (especially components of the Standard Unit Treatment Value Assumption, SUTVA), and offers three strategic responses researchers might adopt: chemosocial minimisation, chemosocial description and chemosocial valorisation. The authors exclude most ketamine trials on the basis that many lack an explicit psychotherapy component and therefore differ in protocol from PAT trials.

This is a conceptual and synthetic analysis rather than an empirical trial or systematic review. The researchers draw on the counterfactual framework of causal inference to explicate assumptions underlying identification of treatment effects, with particular attention to the Standard Unit Treatment Value Assumption (SUTVA), whose two components are consistency (treatments are well defined and uniform) and no-interference (one participant's treatment does not affect another's outcome). They combine this formal causal vocabulary with anthropological and qualitative literature, notably the concept of chemosociality, to theorise social processes that can generate trial-level interference. Rather than reporting a formal evidence search strategy, the paper illustrates arguments with examples and figures drawn from existing PAT trials, qualitative studies of participant experience, regulatory guidance and media events. The extracted text reports counts and timelines of RCTs over several periods (e.g. five RCTs before 2015 totalling 62 patients; seven RCTs in 2016–2020 totalling 193 participants; six RCTs in the past two years amounting to 514 participants) and cites trial durations (for example, a Phase 3 MDMA for PTSD trial took 1 year 9 months from first recruitment to last visit; a Phase 2b psilocybin trial took about 2.5 years). These empirical touchpoints are used illustratively to show the temporal and social opportunity space for dark loops to form. Analytically, the authors use formal causal concepts (average treatment effect, mediation, interference types such as direct interference and contagion) to explain how unblinding and expectancy can act as mediators that amplify spillover. They also integrate findings from qualitative work and regulatory commentary to unpack centre effects, recruitment practices and participant networks, then propose practical responses framed as three positions for trial design and reporting.

The central argument is that PAT trials are especially susceptible to systematic breaches of SUTVA because psychedelics tend to cultivate strong social formations (chemosocialities) and because unblinding in these trials is common. From the causal-inference perspective, violations of the consistency and no-interference assumptions complicate identification of average treatment effects and can bias statistical inference when independence of data points is lost. Measured evidence and illustrative examples presented in the paper include: growth in the number and size of PAT trials (numbers quoted above), long trial durations that span months to years (providing temporal opportunities for interaction between early and late participants), and high rates of unblinding that turn expectancy from a covariate into a potential mediator of treatment effect. The authors outline two interference mechanisms particularly relevant to PAT: direct interference (one participant's treatment affecting another's outcome independently) and contagion (one person's treatment outcome affecting another's outcome). They argue that participant communication, community recruitment, participant-formed groups, centre-specific cultures, and public testimonials all serve as pathways for these interference effects. Specific mechanisms described as producing dark loops include: recruitment from clustered communities and online networks that create pre-existing ties between participants; formation of new participant–community groups and ongoing contact with trial staff that extend effects beyond the trial setting; centre effects arising from site culture, researcher personae or therapeutic allegiances; and widespread patient testimonies and media coverage that shape expectations for current and future participants. The authors note that recruitment materials and selection procedures can preferentially attract participants with positive expectancies, amplifying selection bias and expectancy effects. Interpreting these phenomena, the paper warns of potential instability in observed effect sizes over time: early trials might show inflated effects due to strong expectancy and chemosocial amplification, whereas effect sizes could attenuate or vary as the field matures, recruitment diversifies and social novelty wanes. This instability has implications for regulators and payers because an intervention approved in one sociocultural moment might show different effectiveness later if social drivers are not accounted for. In response, the authors offer three pragmatic positions with concrete measures. Chemosocial minimisation aims to attenuate dark loops through measures such as discouraging participant attendance at integration groups during follow-up, recruiting from existing clinical services rather than open community sampling, improving blinding (including investigating active control conditions), and manualising protocols across sites to reduce centre effects. Chemosocial description proposes systematically documenting chemosocial processes: thoroughly measuring expectations (and distinguishing acute-effect expectations from expectations about outcome and hope), funding qualitative and ethnographic work around trial matrices, involving former participants in interpretation, reporting recruitment materials and timelines, and convening post-study reflections on selection processes. Chemosocial valorisation suggests embracing and optimising chemosocial effects where ethically appropriate, for example via participatory action research, iterative adaptation of protocols, building guardrails with ethical boards and community advisory bodies, and treating emergent group-support processes as part of therapeutic effect rather than noise.

Noorani and colleagues interpret their synthesis as a formalisation of what they call the Pollan Effect—the idea that popularisation and social movement dynamics reshape the interventions studied in PAT trials. They contend that because psychedelics readily foster chemosocialities and because unblinding is common, RCT assumptions central to causal attribution are at risk. From this vantage point, the observed outcomes in many PAT trials may reflect a composite of pharmacological effects, psychotherapy, expectancy and social feedback loops that are currently under-documented. The authors position their argument relative to prior work by acknowledging calls to avoid both overhype and overscepticism in the field, but they add that more explicit attention must be paid to how social context and media-driven expectations feed back into trials. They stress that these issues are not unique to psychedelics but are especially pronounced here because of the substances' socialising effects and the burgeoning public interest. Key limitations and uncertainties acknowledged include the paper's conceptual rather than empirical remit: the authors do not present a systematic empirical measurement of dark loops across trials, nor do they provide definitive quantification of how much chemosocial processes contribute to observed effects. They recognise practical and ethical challenges to the minimisation strategy (for example, restricting participants' access to community support may be harmful and difficult to enforce), and they note that it remains uncertain whether effective blinding for psychedelic substances can be achieved. The authors also explicitly leave unanswered the deeper question of why psychedelics are so chemosocial, citing space constraints and the complexity of providing explanations that are neither culturally nor pharmacologically reductive. Implications discussed include a call for trialists, regulators and health economists to account for chemosocial drivers when interpreting efficacy estimates and making policy decisions. The paper recommends that future research incorporate mixed methods (quantitative expectancy measurement, qualitative ethnography and participant involvement) to document dark loops, and suggests comparative work charting how an indication develops across multiple trials as a test case for their argument. Finally, the authors present three pragmatic stances (minimisation, description, valorisation) as non-exclusive routes that trial teams might adopt depending on epistemic priorities and ethical considerations.

The authors conclude that RCTs operate within real-world social ecosystems and that failing to acknowledge the chemosocial dynamics around PAT risks both misinterpreting trial effect sizes and missing opportunities to study important social mechanisms of change. They reiterate that breaches of SUTVA and rampant unblinding threaten the traditional validity claims of treatment-efficacy assessments in PAT RCTs, while also emphasising that these same dark loops could contain ethically and clinically valuable processes. The paper closes by urging that trials either minimise, document or intentionally harness these chemosocial phenomena so that both epistemic clarity and participant welfare are better served as the field matures.