In an era marked by the digitalization of psychiatry, American researcher Ben Berners-Lee has conducted an observational study on a digital mental health trial, offering an intriguing glimpse into the complexities and contradictions of digital psychiatry. His study, titled āReconciling healthism and techno-solutionism: An observational study of a digital mental health trial,ā was published in the journal Sociology of Health & Illness.
The digital mental health trial in question uses a smart system to draw correlations between data acquired from wearable devices and mood self-reports, further offering behavior modification recommendations for mood improvement. According to Berners-Lee, descriptions of the trial emphasize the strong role of the intrinsically motivated, responsible participant and the empowering influence of the machine learning (ML)-based technology.
Interestingly, this dynamic extends what Berners-Lee calls the āneoliberal paradox,ā where freedom, in order to be preserved, must be constantly adjusted through discipline. As a result of the contradictory nature of this relationship, disagreements about the balance of agency between the objective machine learning system and the empowered participant are frequent among laboratory members.
Neoliberalism and surveillance have been steady companions in the field of mental healthcare. With mental health apps becoming commonplace, the urgency to understand their theoretical and practical underpinnings intensifies.
Ben Berners-Lee unveils the neoliberal ideology that fuels these apps by providing an in-depth account of a digital mental health trial. The study offers insight into the creation, execution, and constant adjustments needed in digital mental health trials to maintain the balance between intention and ideology.
āIn advance of the trial, it was characterized as data-driven, evidence-based, and highly personalizedāa way of helping participants understand themselves through data and thus learn new strategies for improving their mood. By this account, the outcome of the intervention would be determined by the combination of two primary factors: the participants themselves, who are characterized as capable and intrinsically motivated, and the ML [machine learning] system, which would empower participants to achieve their mental health potential.ā
Before observing the digital mental health trial, Berners-Lee conducted a literature review that emphasized the current critical work in translational psychiatry, digital phenotyping, and surveillance capitalism. Guided by this literature, Berners-Lee picks āhealthismā and ātechno-solutionismā to describe the phenomena he saw during his research.
āHealth discourses that emphasize the responsible individual have been discussed by sociologists under the banner of āhealthism,ā and those that focus on the promising epistemological power of new technology are often called ātechno-solutionism.āā
Berners-Lee acknowledges that health consciousness is a useful lens to understand the history of health technology. However, he notes that it risks painting an image of an ideal neoliberal subjectāa controlling rather than empowering presence for the user.
Berners-Lee gathered observational data for exploration over nine months using a trace ethnography method. He observed the laboratory practices of a translational psychiatry lab, which at the trialās outset was described as a āpersonalized, evidence-based, data-driven, machine-learning-based intervention for mental health in mildly to moderately depressed patients.ā
The trial involved several stages for each participant:
- Screening for eligibility in terms of age and exclusionary medical conditions.
- Completing an EEG test.
- Getting fitted with a smartwatch and downloading a mental health application.
āThe smartwatch will automatically monitor steps and other movements using onboard gyroscopes and accelerometers. The mobile application will elicit participant self-assessments of mood, exercise, and diet. These prompts by the mobile app are called ecological momentary assessments (EMAs).ā
In the following 15-day phase, participants passively generated biometric data using the smartwatch and actively reported data concerning their mood, exercise, and diet. This data was sent to a ādata team,ā which developed a machine learning prediction model. Following this, the team selected a Personal Wellness Plan (PWP) to promote optimal mental health.
āThe participant will receive āPWP guidanceā on how to interpret the ML modelās output and implement the proposed behavioral modification plan. Before this process can begin, however, the data team first will explain the ML modelās outputs to the PIs, and the PIs will then explain them to PWP guides. These guides will prepare for guidance the process by completing a series of trainings that simulate the guidance situation. Guidance sessions start with the selection of and introduction to the participantās assigned plan: sleep, diet, positivity or exercise. After familiarising the participant with the overview of the plan and the relevant handouts, subsequent coaching sessions consist of setting weekly goals and evaluating the previous weekās behaviours.ā
The personalization of the wellness plan aimed to stimulate āintrinsic motivation,ā according to Berners-Lee. If thereās a lack of inherent buy-in, the appās purpose is defeated. The app is designed to translate everyday life data into quantifiable figures, and if the participant doesnāt trust the feedback as relevant or personalized, they wonāt implement the intervention. Creating a meaningful intervention becomes a joint effortāif one party disagrees, the output and outcomes become meaningless.
āThe two components, the individual participant and insightful data system that empowers that participant combine to constitute the force of the intervention as it is described in formal accounts. The former component, the participants themselves, falls into the domain of psychiatric expertise, which includes theories of motivation, mood, and well-being that are specific to the experience of the individual. The latter component, the ML system, falls into the domain of computer science and computational modeling.ā
However, how the app resolves potential friction becomes crucial. During the trial, Berners-Lee discovered:
āā¦The two PIs, however, one with a background in psychiatry and the other with expertise in neuroscience and computational modeling, do not entirely agree on the balance of the two key components. In one participantās case, transcribed as Excerpt 1, the algorithm found that variables relating to both diet and social connection had a relationship to mood. When the two PIs are preparing the PWP guide by picking a plan and instructing them on how to propose the plan to the participant, they come to a disagreement. This disagreement relates to the question of the comparative importance of data insights and intrinsic participant motivation.ā
Disagreements and mismatches are not uncommon between mental health and computational machine learning, as seen in other healthcare wearable trials. Itās a delicate balance: the participant must be free enough to follow the appās advice while also recognizing their inability to make rational and healthy choices independently.
One of the most intriguing findings of the study is the crucial yet undervalued role of guides. These are the individuals who assist participants in interpreting the ML outputs and implementing system recommendations. While these guides play a significant role in the trial, they are often relegated to the sidelines and depicted as replaceable in formal accounts.
Berners-Leeās study shows that the work of guides is generally relegated to the technological aspect of the trial, which affects the manner in which they describe and perform their work. This finding, Berners-Lee argues, sheds light on how new forms of human labor can be dismissed or rendered invisible in digital health interventions.
Beyond its contribution to sociological methods and theories, Berners-Leeās study offers valuable insights for study design and computational system design. The study advocates for a co-design approach for health technology that focuses on participant-guide interactions. This would accommodate shifts in relationships between people, professions, and technologies and elicit valuable participant or patient feedback in the design process.
In essence, Berners-Leeās research underlines the urgent need for further critical social scientific research in the field of digital phenotyping. It calls for a more comprehensive understanding of cognitive behavioral therapyās role in the work of guides and lab members and a deeper exploration of participantsā lived experiences that are often excluded from the formal accounts of such trials.
****
BernersāLee, B. (2023). Reconciling healthism and technoāsolutionism: An observational study of a digital mental health trial.Ā Sociology of Health & Illness. (Link)
Change comes when people fellowship and love one another. This is better than Big brother watching for bad actors to intervene.
Report comment
As I recall phenotyping was a concept that meant the observables of gene action in a given environment. Like an A or B blood type was “above” having AA, or AO, or BB or BO blood type GENES, which is not the case of any psychological or psychiatric observable. If it were, it would fall into the really organic diseases, like neurological, metabolical or endocrine.
Try the heterozygous falciform anemia description and history for an enlightenment. To me sounds like a missuse of the word phenotype since it at least disconects it from it’s original context and meaning. To me, at least an equivocation. There are more appropiate and accurate words like observable, variable, measure, etc. Maybe they were trying to avoid the word symptom? As in psychiatric symptom?. Like avoiding “degenerate” when using the word “sensible” for histeria?.
And, rhetorically, it brings back without context the old missunderstood theory of evolution as survival of the fittest. With a fit band!?…
āpersonalized, evidence-based, data-driven, machine-learning-based intervention for mental health in mildly to moderately depressed patients.ā That to me sounds illegal. You cannot do an intervention that has NOT been proved safe and effective in A particular individual, that breaks at least the “state of the art” parameter or judging criterion. Not making judgements nor legal or medical advice. It’s called personalized evidence based medicine. How did the study make sure that for each individual the interventions were safe and effective for each and everyone of them, and of course the aggregate? BEFORE actually implementing them, particularly, when there are no animal models to try that before going human?.
“ecological momentary assessments (EMAs).ā I am sure the environment of each individual was not meassured, before, during or after the whatever was done. Calling the planet ecological does not give license to calling an intervention on it ecological. My gut has an ecology, but that does not give me the leeway to call my diarrhea anthropogenic global warming…
“Following this, the team selected a Personal Wellness Plan (PWP) to promote optimal mental health.” And the evidence, scientific that is, before putting forth that is what exactly? Now, how was the optimum meassured in each individual? Was it like the intersection of risk and return in your assets investment?. Like the risk/benefit economic frontier depicted in MANY economics textbooks?.
“behavioral modification plan”, yeah, it sounds big brotherly…
“” āThe participant will receive āPWP guidanceā on how to interpret the ML modelās output and implement the proposed behavioral modification plan. Before this process can begin, however, the data team first will explain the ML modelās outputs to the PIs, and the PIs will then explain them to PWP guides.” ” That to me sounds like the young Lister from Red Dwarf would have said: Cryptofascist.
“insightful data system”. There are no data systems that give insight into the human mind. The human mind, last time I checked, scientifically, even mathematically, is A BLACK BOX. No insight there, according to mathematics and COMPUTATION science.
“theories of motivation, mood, and well-being” as far as I recall are not psychiatric, are PSYCHOLOGICAL.
“…The former component, the participants themselves, falls into the domain of psychiatric expertise, which includes theories of motivation, mood, and well-being that are specific to the experience of the individual. The latter component, the ML system, falls into the domain of computer science and computational modeling.ā And the missing part, at least the interaction between the two of them falls, to me, into quackery. Psychoanalysis and psychodynamy have been called pseudoscience, and they most probably are, to state they are beyond ANY doubt, reasonable that is, requires knowing how they work in nature, same as ghosts, gods, angels, demons, and sadly aliens, since I do want to believe, still today, many decades past. It is to me the epistemic qunitessential for being addressable with the scientific method: falsifiability.
Reusing their language with a ML addon still sounds to me pseudoscientic.
“Itās a delicate balance: the participant must be free enough to follow the appās advice while also recognizing their inability to make rational and healthy choices independently.” Can you freely and “independently” follow an advice on which you cannot assign a value of rationality or healthiness?. That’s called, as far as I recall something aking to faith, or uuggh! dogma, not freedom, historicaly and ethymologically. Knowledge is part of freedom, paraphrasing Yogi Berra, particularly about the future.
Not trying to be offensive nor dismissive, among others, but this analysis and summary sounds to me of a pseudoscientific, uncritical, ignorant, and of the vaginal candles and pessaries type. Sorry, but I already stated my case. š
And as an apology for any inconvinience: I am an old folk, of the skepticamp, from a superstitious country, that practiced impressively but poorly David Copperfield’s feats at a young age (with a profit!, hence my touchy issues) and was always a fan of James Randi. Not trying to be hurtfull nor offensive, just your average Krieger bot: “Too damn sensitive”.
Report comment
Mental health smart watches could be a big fad in a few years just like mindfullnrss apps were a few years ago. Meanwhile prescriotions continue to rise as does detention under MH laws in the UK.
Report comment
Thank you for this article. I especially appreciate the inclusion of neoliberalism as a root cause of many problems. I worry that many among us donāt realize how much monied interests now rule our systems, including in mental health. You are explaining and exposing us to these often hidden structures.
Report comment