Sage Journals: Discover world-class research

Abstract

This article critically examines the emergence and uses of digital phenotyping in contemporary psychiatry. From an analysis of its discourses and practices, we show that digital phenotyping diffusion is directly related to its promise to solve some of the major impasses of the so-called "neuro-turn" in contemporary psychiatry. However, more than a new tool to address old objects of pre-digital psychiatry, we consider digital phenotyping as participating from a new onto-epistemological matrix, the “neuro-digital complex,” which entails the redefinition of psychiatric objects (e.g., brain and mind), diagnostic categories and procedures, subjectivities (e.g., users of mental health apps), and the emergence of a new regime of truth which promises to reveal the neuropsychological core at the individual scale. Despite this techno-utopia, digital phenotyping does not produce neutral mirrors for self-knowledge. We show that it resorts to population statistics, grounded truth data sets built with pre-digital neuropsychological assumptions, and human categorization processes. Nevertheless, we propose not to approach this gap as a misleading ideological fact but to emphasize its productive possibilities. From this perspective, the gap becomes the measure between whom we think we are and who we really are, working as a guide to conduct our lives in neuropsychological terms. Thus, we conclude that, rather than providing personalized diagnoses and treatments, digital phenotyping produces individualized pathways to normalization and neuropsychologization.

Keywords

Digital phenotyping self-tracking mHealth digital psychiatry therapeutic cultures science and technology studies

Introduction

In September 2018, Thomas Insel,¹ a psychiatrist and former head of the US National Institute of Mental Health (NIMH), speculated about the future of his discipline:

In 2050, when psychiatrists look back at the first two decades of the 21st century, what will they recognize as having the greatest impact? No doubt the revolution in genomics […] and the revolution in neuroscience […] will be considered important. But perhaps the revolution in technology and information science will prove more consequential for global mental health (Insel, 2018: 276).

Insel's futuristic imagination condenses two of the most relevant changes in the past 30 years in psychiatry and psychological discourses and practices. The first shift has been called the “neuro-turn,” which is linked to the process of “cerebralization” of the mind that took place since the mid-20th century (Littlefield and Johnson, 2012; Rose and Abi-Rached, 2013; Vidal and Ortega, 2017). The second more recent shift, which we call the “neuro-digital complex,” is related to the impact that digital technologies and artificial intelligence (AI) are beginning to have on neuroscientific and psychiatric discourses, practices, and promises (Mohr et al. 2017; Martin-Key et al., 2021). The rapid spread of internet-connected devices and various digital sensors embedded in everyday life, such as smartphones, smartwatches, along with fitbits and other wearables, have made it possible to collect and analyze a continuous data stream of individuals’ behaviors, psychological states, and environments, forming a picture of peoples lived experiences (Mohr et al., 2020).

In this new cultural, scientific, and technological context, the notion of “digital phenotype” has recently emerged (Jain et al., 2015). Using the traditional conception of phenotype as the physical and behavioral expression of an organism's genotype—size, metabolic activities, and patterns of movements—and incorporating Richard Dawkins’ (1983) concept of “extended phenotype,” Jain et al. (2015) enlarged this concept to include human–computer interactions. The basic assumption is that the AI analysis of data collected by digital devices embedded in everyday life—such as geolocation, step counters, social interactions, typing speed, biometrics, and physiological parameters, among others—provides valuable information on the objective foundations of individual mental health. This enables the possibility of building “digital phenotypes,” that is, patterns of individuals’ mental health (Insel, 2018; Piau et al., 2019), with the aim of “detecting or predicting the onset of mental illness and rapidly disseminating effective and affordable care to those who need it” (Tekin, 2020: 450).

The notion of digital phenotype has been widely circulated in recent years. Related concepts quickly began to emerge in the field of psychiatry, such as “digital phenotyping” (DP) (Torous et al., 2016)—which emphasizes its dynamic nature over the more static construct of digital phenotype (Onnela and Rauch, 2016)—or the notion of “digital biomarkers,” which focuses on the possibility of using data collected from digital devices to distinguish between symptomatically indistinct mental health illnesses and to detect and treat pathogenic processes even at preclinical stages (Haag, 2007; Hidalgo-Mazzei et al., 2018).

DP is currently one of the most used terms in publications listed in PubMed, spreading rapidly in the healthcare industry (Mohr et al., 2020). Currently, there are open-source platforms, such as Harvard's “Beiwe” (Onnela, n.d.) or “HOPES” (Health Outcomes through Positive Engagement and Self-Empowerment) (Wang et al., 2021), which provide researchers with a plethora of personal data collected in the wild. Thus, much research is being done about DP in a wide range of areas, such as identifying, tracking, and predicting treatment responses for mood and anxiety disorders (Rohani et al., 2018; Zulueta et al., 2018; Jacobson et al., 2019; Faurholt-Jepsen et al., 2020; Saccaro et al., 2021); assessing the risk for psychotic episodes (Benoit et al., 2020; Henson et al., 2020); identifying the risk of suicide (Vahabzadeh et al., 2016; Kleiman et al., 2018); and tracking and modeling subjective wellbeing (Rhim et al., 2020), among many other areas of development and innovation.

However, the information produced with digital data collected in our daily lives is transforming psychiatric and neuroscientific practices and pervading popular cultures. The use of mobile wireless technologies for public health, or “mHealth” (WHO, 2018), has become a vast industry. In this context, different start-ups have built smartphone applications such as Mindstrong (n.d.a) and Ginger (n.d.) which use DP. Based on the 24/7 uninterrupted capture of personal data, these apps produce real-time mental health charts, enabling new ways of understanding, assessing, and managing mental health. Thus, while DP transforms the way science understands and approaches mental health, this technology is enabling new everyday therapeutic cultures (Nehring et al., 2020) and “technologies of the self” (Foucault, 1982) through which individuals act on themselves and conduct their lives.

Although there has been a proliferation of critical analyses of digitalization in psychiatry (Pickersgill, 2019) and more recently on DP (Birk et al., 2021), much research remains to be done from a social science perspective. Some of these critical studies have explored DP's ethical and epistemological limitations (Loi, 2019; Lucivero and Hallowell, 2021); others have critically analyzed how DP reconfigures some traditional sociological categories, such as social interaction or social determinants of health (Birk and Samuel, 2020; Rowe, 2021). Meanwhile, other research has focused on the consequences and tensions associated with DP's promise of providing individualized ways to assess and intervene in mental health (Baumgartner, 2021).

This article proposes a critical approach to DP. To reconstruct its discourses, practices, and promises, we carried out a non-systematic search in different databases (Google Scholar, PubMed, and Web of Science), and we analyzed several systematic or narrative reviews on DP published between 2018 and April 2022 (Rohani et al., 2018; Baumeister and Montag, 2019; Piau et al., 2019; Liang et al., 2019; Spinazze et al., 2019; Benoit et al., 2020; Melcher et al., 2020; Martin-Key et al., 2021; Camacho et al., 2021; Mendes et al., 2022). Moreover, to analyze the non-scientific discourses and practices, we searched the web for apps and platforms that claimed to use DP. This review provided a sufficiently robust body of literature to account for DP's reasoning style, discourses, and practices in psychiatry. However, we took a different path to analyze this material than previous sociocritical research. As Germán Berríos (2002) stated, a proper analysis of psychopathological concepts and categories should consider the fact that they “only have meaning and value within a given period and cannot be compared with events from different episteme” (Berríos, 2002: 2). Thus, instead of making a “trans-epistemic” analysis, such as asking whether digital data capture “the truth” (or not) of pre-digital mental disorders (Birk and Samuel, 2020), this paper focuses on the productive effects of DP. This means that we analyze it as an assembly of different human and non-human elements—discourses, practices, and technologies—which produce a specific neuro-digital “regime of truth” (Foucault, 1976). Following Karen Barad (2003, 2007), for whom practices of knowing and being are always mutually implicated, we approach DP as a new onto-epistemological matrix that works as a “surface of emergence” (Foucault, 1972: 41) of particular knowledges, subjects, and objects.

Notwithstanding, we consider this matrix not as a worldview, nor as a transcendental form that constrains knowledge and beings, but as a set of enculturated practices, a “modus operandi that authorizes new facts to emerge” (Latour, 2006: 29, translation is ours). As Bruno Latour (2004) states, to emphasize the onto-epistemological productive effects of DP is not a critique that debunks but one that assembles; rather than lifting “the rugs from under the feet of the naïve believers” the article seeks to provide new onto-epistemological “arenas in which to gather” (Latour, 2004: 246). Therefore, this paper contributes to producing a new “arena” for future critical research, to think not only against but also with DP, visualizing its specificities and how the neuro-digital complex transforms some of the central questions of pre-digital psychiatry. Considering its onto-epistemological specificities does not prevent but enables the possibility of producing well-grounded comparisons, discussions, and dialogues between pre-digital and digital psychiatry. Within this broad field of analysis, we focus on DP's promise to deliver tailored diagnoses and interventions in psychiatry, showing that rather than failing to capture pre-digital mental health categories at the individual scale, DP enables new personalized pathways to normalization and neuropsychologization.

Neuro-turn impasses and the background of digital phenotyping

“As long as the research community takes the DSM to be a bible, we’ll never make progress […] People think that everything has to match DSM criteria, but you know what? Biology never read that book” (Insel, 2013: n.p.).

The neuro-turn encompasses scientific and popular cultures that share the conception that the truth about who we are lies in our “neurochemical self” (Rose, 2003). This contemporary perspective contrasts with subjectivists’ traditions for which the mind could only be approached through the first-person perspective, that is, verbal self-reports analyzed by clinicians (Kozak and Cuthbert, 2016). On the other hand, following the perspective inaugurated by behaviorism at the beginning of the 20th century, the neuro-turn seeks objective biological evidence to explain the mind, avoiding the unreliability of the first-person perspective (Taschereau-Dumouchel et al., 2022). Since multiple biological events can produce similar subjective experiences, and similar biological mechanisms can produce heterogeneous self-reported symptoms, clinical phenotypes became “fictive” entities without biological groundings (Kozak and Cuthbert, 2016; Peterson, 2020). Within this neuro viewpoint, the truth about who we are has migrated from the first to the third-person perspective of brain waves, eye movements, or neuroimages (Dennett, 1991). In other words, the neuro-turn seeks “to allow ‘the power of biology,’ and not the experience of the patient, the symptoms of the disorder, or the judgment of the clinician, to delineate the disease” (Rose, 2019: 86).

A clear example of the impact that the neuro-turn has produced in contemporary psychiatry is the strong criticism received by the DSM-5 (APA, 2013), because its diagnostic categories, based on clinical consensus, could not be aligned with hard biological evidence (Insel, 2010: 748). As David Kupfer (2013), the chairman of the DSM-5 task force, confessed:

The problem that we’ve had in dealing with the data that we’ve had over the five to 10 years since we began the revision process of DSM-5 is a failure of our neuroscience and biology to give us the level of diagnostic criteria, a level of sensitivity and specificity that we would be able to introduce into the diagnostic manual. (Kupfer, 2013: n.p.)

In this context, in 2009, the NIMH started the “Research Domain Criteria” (RDoC) initiative (Cuthbert, 2014). One of this project's main objectives is to identify mental disorders biomarkers, which would replace current clinical diagnoses and psychiatric categories (Kapur et al., 2012: Vilar et al., 2019). Therefore, basic sciences—genetics, neuroscience, and behavioral science—should serve as the starting point for understanding clinical phenomena (Cuthbert and Insel, 2013). However, self-reports and clinical observation are still one of RDoC's units of analysis, but without biological evidence to support them, they are considered “folk-psychology” (Kozak and Cuthbert, 2016: 292). Therefore, rather than eliminating clinical evidence, RDoC seeks to find the psychobiological mediators which could provide a solid and smooth causal connection between the biomarkers of mental illness and their clinical phenotypes.

However, finding those brain-mind mediators has become more of a scientific utopia than an achievable project (Patrick and Hajcak, 2016; Rose, 2019; Peterson, 2020). The first obstacle found by RDoC is linked to the impossibility of producing objective clinical phenotypes that could be translated into biological terms. Many psychiatric disorders impact cognitive functioning; thus: “relying on subject self-report for data on symptoms, behavior and even physiology may be unreliable” (Torous et al., 2017: n.p.). Furthermore, even if the subject does not present any cognitive impairments, the lack of granularity in self-reported symptoms became an insurmountable obstacle for RDoC (Hidalgo-Mazzei et al., 2018).

The second difficulty was associated with the data collection settings. Since “observable phenotypes of psychiatric illnesses are temporally dynamic, environmentally influenced and socially modulated” (Torous et al., 2017: n.p.), scientific and artificial settings make it difficult to study them. Whether in a laboratory or a psychiatrist's office, the discontinuity of these environments with people's daily lives can change how mental illnesses manifest.

Once again, Thomas Insel's statements and biography are illustrative of how the legitimacy of the neuro-digital complex is directly linked with the disappointment produced by the unfulfilled promises of the neuro-turn:

[I] spent 13 years at NIMH really pushing on the neuroscience and genetics of mental disorders, and when I look back on that I realize that while I think I succeeded at getting lots of really cool papers published by cool scientists at fairly large costs—I think $20 billion—I don't think we moved the needle in reducing suicide, reducing hospitalizations, improving recovery for the tens of millions of people who have mental illness. (Insel, 2017a)

Thus, it is not surprising that Insel left the NIMH in 2015 to launch a mental health program within Verily, Google's life sciences group (Ledford, 2015). Then, in 2017, he left Verily to cofound the startup “Mindstrong” with Paul Dagum and Richard Klausner (Insel, 2017b). In this new digital context, instead of neuroimages and genomics, Insel focused on using AI to analyze patterns of interactions with smartphone apps. For Insel (2017c, 2018), the impasses of the neuro-turn created the need for new tools, such as DP, which could capture phenotypes of psychiatric illness in individuals’ real-world environments in real-time, for long periods, and with minimal subjective bias.

The neuro-digital complex: The digitalization of the mind and the de-substantialization of the brain

The almost infinite amount of data produced by the digital devices integrated into our daily lives suggest solutions to some of the major impasses left open by the neuro-turn and the RDoC project (Hidalgo-Mazzei et al., 2018). Consequently, in the last decade, cutting-edge research in neuroscience and psychiatry has increasingly focused on AI and big data analytics (Richards et al., 2019; Huys et al., 2021).

However, the process of “datafication” (Mayer-Schöngerger and Cukier, 2017) of mental disorders is not new. Psychometric instruments began to be used in psychiatric and psychological research protocols starting in the 1930s (Danziger, 1997; Demazeux, 2019). Meanwhile, the use of computers for creating brain models and analogies began in the mid-20th century (Cobb, 2020). For example, in the late 1960s, Robert Spitzer, president of the DSM-III Task Force, carried out a research program that sought to produce computerized diagnoses to avoid the biases introduced by expert and patient subjectivities (Spitzer and Endicott, 1968). Ecological strategies for measuring mental health are not new either. We can trace them back to at least the 1980s, when different methods—such as the “Experience Sampling Method” (ESM) (Csikszentmihalyi and Larson, 1987) and the “Ecological Momentary Assessment” (EMA) (Stone and Shiffman, 1994)—began to be used to analyze symptoms and behaviors in patients’ natural environments.

Thus, to understand the novelty and specificity of DP, one should analyze it in the broader context of the new onto-epistemic assumptions of the neuro-digital complex. From this viewpoint, one crucial aspect is the distinction between “active data” and “passive data” (Onnela and Rauch, 2016). Active data requires the subject to voluntarily participate in data production, like answering a questionnaire or reporting his/her feelings. Previous ecological assessment strategies depended on this kind of information, a subjective interpretation of reality.

Notwithstanding, the spread of portable and wearable digital technologies has created a new form of data:

Human–computer interaction measures not what you type but how you type. Subtle aspects of typing and scrolling, such as the latency between space and character or the interval between scroll and click, are surprisingly good surrogates for cognitive traits and affective states. (Insel, 2018: 276)

These “passive” data are the unintended “footprints” left by people when they use wearable digital devices or interact with software (Bidargaddi et al., 2016). Rouvroy and Berns (2013), note that passive data are more “abandoned than ceded, traces left and not transmitted” (Rouvroy and Berns, 2013: 169, translation is ours). In other words, even if the subject is conscious of their interaction with wearable devices, they have sensors that can capture the most subtle micro-gestures, decomposing the unity of the subject into involuntary, sub-semantic, and quantifiable components such as voice tone, cardiac rate, and steps number, among others.

Therefore, if traditionally clinical phenotypes depended on self-reports and experts’ observations in artificial settings, the first onto-epistemological assumption of the neuro-digital complex is that they can now be fully digitalized, quantified, and, hence, computerized. The algorithmic analysis of passive data promises to bypass the subjectivity of patients and clinicians and the artificiality of scientific settings, becoming the new “digital truth serum” (Stephens-Davidowitz, 2017). Under these conditions, it seems the mind has finally become accessible to the scientific gaze: “After 40 years of psychiatry becoming more mindless than brainless, perhaps digital phenotyping will help the pendulum swing back toward a fresh look at behavior, cognition, and mood” (Insel, 2017c: 1216). The “digital footprints” left by our interaction with wearable devices seem to provide a new scientific way of directly approaching the real subject, avoiding the need for other technological mediators, such as brain images.

The second onto-epistemological assumption relates to a parallel process of datafication and de-substantialization of the brain. Since the beginning of the computer age and the cybernetic imaginaries in the1950s, the brain began to be problematized as computable data and emergent patterns (Terranova, 2004; Cobb, 2020). According to Katherine Hayles (1999), this process entailed a shift from the traditional presence/absence ontology to a pattern/randomness one. In this context, scientists began to construct analogies and metaphors about how the brain and these new machines processed information (Cobb, 2020). However, since the 1980s, linked to the developments of deep learning technologies, another version of the brain-computer metaphor began to emerge, which saw the brain as a series of emergent patterns that could be analyzed with AI (Cobb, 2020: 261). As Rafael Yuste (2015) states, the “neuro-doctrine”—the conception of the neuron as the functional unit of the nervous system—that prevailed for over a century is now beginning to be replaced by “neural network” models of the brain. That is, contrary to the neuro-chemical brain—which was understood and treated as a material external reality that could be segmented into its components and studied with neuroimaging (Vidal and Ortega, 2017: 63)—the new neuro-digital brain is understood as a complex system whose major goal is to produce emergent functional patterns (Yuste, 2015).

The digitalization of the mind and the de-substantialization of the brain reduced them to computable data that could be analyzed with AI. Understood as such, DP can produce new hybrid patterns of mental health, combining passive and active data with hard biological evidence. Consider, for example, how the Mindstrong app presents its methodology:

Research volunteers completed extensive neuropsychological testing, clinical assessments of mood and cognition, and, in some cases, neuroimaging with fMRI. The results revealed a set of digital biomarkers from human-smartphone interactions that correlate highly with select cognitive measures, mood state, and brain connectivity. These interactions were comprised of taps, swipes and other touchscreen activities which were completely content-free and not reliant on capturing personal information. (Minstrong, n.d.b)

As long as DP conceives the brain and the mind as a set of emergent patterns built with computable data, rather than “bridging the gap” between them, it dissolves the dichotomy.

The third characteristic of the neuro-digital complex is the redefinition of some major psychiatric diagnoses and intervention features. As suggested by Yarkoni and Westfall (2017), one of the main goals of pre-digital psychiatry was to explain the underlying causes of mental illness and to predict its triggering. However, they argue, “although explanation and prediction may be philosophically compatible, there are good reasons to think that they are often in statistical and pragmatic tension with one another” (Yarkoni and Westfall, 2017: 1100, italics in the original). In other words, due to the nonlinear complexity of mental illness, contemporary psychiatry must choose between understanding or predicting. The consequences of this tension are vast and not always obvious. For example, as we have noted, DP often combines active data with passive data, which appears to be a necessary requirement for improving the predictive power of DP (Currey and Torous, 2022). Yet, to be intelligible for algorithms and AI, active data must be treated as any other kind of data, that is, as a sub-semantic chunk of digital information which can be combined with passive or genomic data, among many others. For example, if self-reported low-mood feelings improve predictability, it does not matter whether they are caused (or not) by a “real” depressive disorder. There is no possible way to know why and how self-reports improve algorithmic prediction levels; it just works. The digital clinician is no longer confronted with an object—spatially localized, discrete, and stable over time—but with processes and patterns of covariances, and its main objective is not to understand the hidden causes of mental disorders but to produce tailored profiles to predict and act even at preclinical stages (Turkheimer et al., 2019; Stanghellini and Leoni, 2020).

Therefore, the novelty and relevance of DP in psychiatry is that it does not really “solve” but actually “dissolve” some of the neuro-turn impasses. Instead of finding the psychobiological mediator between clinical phenotypes and hard biological evidence, DP redefines what the mind and the brain are. Regarding the former, self-reporting and the clinical gaze have been either replaced by passive data gathering and/or treated as such. Concerning the brain, AI is not just a more powerful tool to explain what neuroimages could not: it redefines the brain in terms of emergent patterns, giving correlations a status like that of causal explanation in the neuro-turn.

“Closing the loop”: The promise of digital phenotyping mental health interventions

DP is a techno-discursive paradigm that provides new frameworks for understanding human actions and new imaginaries of who we are, which redefine how mental disorders and psychiatric interventions should be understood. One of the main possibilities introduced by DP is building mental health categories in real-time, continuously evolving with new data inputs. This is seen as producing individualized diagnoses and tailored interventions—for treatment or prevention—minimizing the elapsed time between them (Huckvale et al., 2019: 6). For the moment, DP is used to inform experts or patients so they can make the best possible mental health decision at the individual scale at the right moment. For example, a promotional video of the mental health app Ginger describes how DP interventions work:

[digital phenotypes are sent] to someone who can act. Whether it's a care manager who's trying to decide which 30 out of 300 patients to call on a given day; whether it's a clinician who wants to use that information to deepen the therapeutic alliance during a session; or whether it's even back to the patients themselves in the form of health tips and other in-app tools. (Ginger.io, 2015: 2:35–2:50)

Therefore, the apps are fed with passive data; then, algorithms build probabilistic patterns about individuals’ states of mind. These are translated into charts given to experts or patients in real-time so they can intervene in the fastest possible way, even at the preclinical stages of a disease.

With access to moment-by-moment individuals’ psychological, behavioral, and physiological data, DP has the potential for automated forms of intervention. These new technologies would depend neither on experts’ judgments nor on the users’ choices but on algorithmic decisions. Ecological Momentary Interventions (EMI), and their more radical expression, Just-in-Time-Adaptative-Interventions (JITAIs), are a suite of different strategies that adapt over time to an individual's changing internal and contextual state with the goal of providing the right type and amount of support at the right moment (Wang and Miller, 2019; Balaskas et al., 2021). However, as Torous and colleagues (2021) have recently shown, this combination of DP with JITAIs is still more of a near-future promise than a current reality. Contemporary JITAIs do not use passive data but still rely on self-reporting strategies.² Yet, they give us the following example of what they would look like working together:

[…] the smartphone may be able to infer low mood in the context of social isolation and offer a relevant intervention, whilst, in another circumstance, it may infer low mood in the context of poor sleep and recommend an alternative intervention. Although in its infancy, using JITAIs to offer ‘closed loop’ mental health interventions is a promising area for future research. (Torous et al., 2021: 320)

The central promise of DP for mental health interventions is to “close the loop” (Huckvale et al., 2019: 6), that is, to shorten the elapsed time between diagnosis and intervention; and to predict, prevent and/or even treat mental disorders at preclinical stages (Torous et al., 2017). Therefore, DP is part of the contemporary “prevailing Weltanschauung for mental health research” (Peterson, 2020: 1279): the promise of bringing “precision medicine” into psychiatry. This new perspective aims to dismiss the coarse granularity of traditional mental health disorder labels to produce personalized diagnostics and treatments for each individual or specific population (Baumgartner, 2021).

The imaginary of “closing the loop” relies on at least two assumptions. The first is that DP can produce hyper-realistic representations of the self with personalized moment-to-moment ecological passive data. For example, some studies point out that we can identify states of loneliness or depressive symptoms in clinical and non-clinical populations through correlating data captured via GPS, accelerometers, calls, SMS, etc. (Rohani et al., 2018; Doryab et al., 2019). The second assumption is that the reduction of the elapsed time between data inputs and interventions would make it possible to act upon the same reality that is being captured. For example, it would be possible to intervene when a person presents suicidal thoughts or behavioral warning signs before the suicidal attempt (Vahabzadeh et al., 2016).

To critically analyze these assumptions, there are at least two options. First, following Birk and Samuel (2020), it is possible to question whether it is true that DP is a technology that allows capturing old mental health problems in a more personalized and precise way than pre-digital psychiatry. Is it possible to infer loneliness, a self-interpreted social situation, from passive data? Or can we presuppose that lack of mobility, as captured by digital devices, necessarily indicates loneliness, social isolation, or depression? Is it possible to predict cognitive impairment or suicidal behavior from voice recording or typing speed on a smartphone? What data and criteria count as a benchmark for “normality”? Likewise, when Torous et al. (2021) imagine that the use of DP with JITAIs could detect that low mood sometimes is associated with poor sleep, while others with social isolation, one may ask if they are really measuring these phenomena. Does this new method address the same old symptoms but on a supposedly more personal scale? One possible answer to these questions is that DP fails to assess “real” pre-digital symptoms because “there are some states that cannot be inferred purely from passively sensed data” (Birk and Samuel, 2020: 8). From this perspective, self-reports, and introspection cannot be surrogated by passive data. Thus, DP becomes a kind of “false consciousness” that deceptively promises to measure the old analog psychiatric categories better. However, the problem with this critical perspective is that it does not consider the onto-epistemological leap between pre-digital and digital psychiatry. From this second perspective, both pre-digital and digital loneliness are not two different species from the same genus but different species from other genera. DP is not a sort of “false consciousness” that fails to capture the old “real” loneliness, but it redefines what it means to be alone, to be with others, and to be tout court.

Therefore, in what follows, we take another critical approach. Instead of confronting DP with pre-digital psychiatric categories, we will analyze it from the “inside,” emphasizing its productive effects. From this viewpoint, rather than showing how DP fails to capture old psychiatric categories, we explore the cultural and technological context that enables the possibility to identify, for example, depression and loneliness with the algorithmic patterns produced with passive data. How is it possible that our personal neuropsychological reality, which was supposed to be the inaccessible hidden core of ourselves, has now been turned into objective mental health charts? What are the normative and performative effects of this new technology? Who are we becoming in the neuro-digital complex era?

Thinking with and against digital phenotyping

As we have seen, one of the main promises of DP is to produce personalized diagnoses and tailored interventions. Pre-digital psychiatry worked either at the N = 1 scale, analyzing intra-individual variability, such as in clinical settings, and/or at the N = All scale, as in epidemiology, harvesting a handful of measurements from thousands of people, analyzing inter-individual variabilities. The use of digital devices in psychiatry has allowed, as never before, to analyze human behavior on a large scale and in fine detail simultaneously, enabling new possibilities to study the mediation processes between them (Markowetz et al., 2014; Schork, 2015; Onnela, 2021).

To find meaningful patterns of mental health at the N = 1 level, algorithms should be trained with “grounded truth” data sets, that is, a “human-based classification, acting as the reference norm against which all machine-based classifications are to be evaluated” (Grosman and Regeluth, 2019: 3). In other words, data must be readied before algorithms can use them, and these processes usually include human decisions about what data are relevant or irrelevant (Gillespie, 2014). Different scholars have shown how this process of manual labeling to train algorithms reproduces and encodes within them all kinds of social inequalities and cultural biases, such as gender or racial prejudices (Nakamura, 2008; Browne, 2015). Therefore, as Green and Svendsen (2021) state, DP information is “inherently comparative and relational, and digital mirrors can be bent or stretched in specific directions, depending on the choice of reference classes” (Green and Svendsen, 2021: 2).

To illustrate these issues, we can resort to Beth Semel's (2021) ethnographic description of the process of developing a smartphone app that purports to analyze minute changes in phone calls’ speech sounds—digital vocal markers—to predict manic or depressive episodes. To build the grounded-truth data set, the research team cross-referenced people's diagnostic, obtained with psychological questionnaires, with their voice tones labeled by people trained to listen “as computers”—paying attention to speech sound and inattention to speech meaning.³ Semel (2021) concludes that just like other forms of work in late capitalism that claim to be the product of autonomous machine labor, DP is, in fact, “a contingent, sociotechnical achievement” (Semel, 2021:17). In other words, digital phenotypes are not just efficient codes but assemblies of different forms of human labor, material resources, and ideological choices (Finn, 2017). This means that they do not capture the actual neuropsychological core of people but rather what psychological questionnaires and laypeople's assumptions consider a mental disorder. And these biases are not innocuous. They produce normative effects on the people being assessed and monitored by DP (Bradstreet, Allan, and Gumley, 2019).

However, these neuropsychological preconceptions and generalizations seem to disappear at the N = 1 scale. From the experts’ viewpoint, passive data promise to give them direct access to the biological foundations of individuals’ observable symptoms (Semel, 2021). On the other hand, app users get the impression that digital phenotypes produce a reified depiction of their inner neuropsychological truth. For example, on Mindstrong's website, there is a testimonial video of a young woman diagnosed with bipolar and anxiety disorders, who describes her experience with the app as follows: “I have this care team with Mindstrong that's going to be there when I need them, and I can see myself with the charts that they provide for you on that app” (Mindstrong, n.d.c: 1:16–1:26, italics are ours). The charts produced with DP methodologies allowed the user to “see herself,” i.e., to see her inner neuropsychological truth. Therefore, there seems to be a gap between the effect of algorithmically personalized truth and how that truth is produced. In other words, digital mirrors are not neutral surfaces for self-knowledge (Vegter et al., 2021), a fact that contrasts with the aseptic experience of individualized truth produced by them. Following Benjamin (2019), it is possible to assert that DP is a technology that reproduces different kinds of neuropsychological questionable assumptions and generalizations but is nevertheless perceived as more objective, neutral, and personalized than pre-digital psychiatric technologies.

How will we understand and analyze this gap between how neuro-digital truth is produced and how it is perceived? Shall we think against DP to pull “the rugs from under the feet of the naïve believers” (Latour, 2004: 246) and show them that DP is a deceptive technology that claims to be neutral and objective while, in fact, it is an ideological artifact? The “deceptive hypothesis” supposes that the same subjects are being exposed to different ways of knowing equivalent objects; that is, it entails the disjunction between the way things are known—epistemology—and what is to be known—ontology. On the other hand, following Karen Barad (2003, 2007), to think with DP implies assuming the inseparability of the onto-epistemological realms. This means that the digital episteme is not a new tool for knowing old things but produces new objects and subjects (Bemme, Brenman, and Semel, 2020). Our main critical purpose is not to show if digital psychiatry says the truth (or not) about certain phenomena but rather to analyze what are the sociotechnical conditions for the emergence of the new personalized neuropsychological truth and which are its main productive effects (Foucault, 2008: 36). In other words, to think with DP entails analyzing the new cultural patterns, subjective processes, and sociotechnical characteristics which relate to the emergence of the neuro-digital truth.

As previous critical analyses have shown, pre-digital therapeutic cultures were modulated by processes of neuropsychologization, that is, by the production of subjects who had to learn to understand and conduct their lives as experts in neuropsychology (Rose, 1998; De Vos, 2013; Nehring et al., 2020). This cultural context should provide “good” patients with at least two essential skills: they should be able to analyze their feelings and behaviors as neuropsychological experts to produce active data and to use neuropsychological discourses as a hermeneutic tool to make sense of diagnoses and treatments. Regarding the latter, just as neuroimages are meaningless without neuropsychological knowledges (Vidal and Ortega, 2017), DP charts still depend on that sort of discourse as their hermeneutic counterpart. In other words, neuropsychological cultures are essential to make data “speak” (Dourish and Gómez Cruz, 2018). However, unlike the therapeutic cultures of the 20th century, app users do not need to behave like neuropsychological experts to produce active data about their mental states. Algorithms and AI have outsourced “the psychologization to the technological: We do not have to psychologize ourselves, this is done by technology” (De Vos, 2020: 239).

Thus, DP's effect of truth partially depends on the unexpired validity of pre-digital neuropsychological cultures and subjectivities and partially on the spread of other kinds of discourses, practices, and promises known as “self-tracking cultures” (Lupton, 2014, 2016). These cultures are linked to the emergence of wearable digital devices, which are used to permanently monitor and measure physiological, psychological, and/or behavioral elements, providing new forms of self-knowledge that nudge people to improve different physical or psychological characteristics. Even though self-tracking practices seem highly individualistic, they are embedded in broader discourses on technology, selfhood, the body, and social relations, which give them meaning and legitimacy (Lupton, 2014). Just as psychoanalysis taught people to recognize themselves in their unconscious formations, contemporary self-tracking cultures make it possible for people to recognize themselves in charts produced with passive data.

The desire to feed applications with passive data and then experience the charts they produce as the truth about ourselves is related to this new cultural context. For example, a self-tracker interviewed for a Washington Post article (Hesse, 2008; quoted by Lupton, 2016: 90) justifies the importance of these practices as follows: “I want to understand the changes that are actually happening [in my life], not just my perceptions of them” (Hesse, 2008: 5). Then the author of the article continues: “Has it really been a month since you last had sex, or does it just feel like that? […] Computers don't lie. People lie” (Hesse, 2008: 5). From this perspective, if you feel depressed or lonely, but your digital phenotype contradicts that, you should question your self-perception.

Therefore, DP's personalized truth would result from the emergence of new subjectivities linked to the contingent gathering of pre-digital neuropsychological cultures with self-tracking practices and the new kind of data produced by digital technologies: the neuro-digital complex. The gap between how DP produces truth and its effect of personalized truth must be reinterpreted more complexly. When analyzed from pre-digital neuropsychological discourses, the gap could be considered a deceptive leap (Birk and Samuel, 2020). DP claims to reveal my depression and loneliness, while as we have shown, it reproduces questionable neuropsychological assumptions and generalizations. However, this gap has new productive effects as well. Self-tracking sociotechnical practices have taught people that digital devices can capture and analyze their individually experienced lives. Consequently, when digital mirrors reflect something that does not entirely coincide with self-perceptions, the distance is reinterpreted as the difference between how we perceive our lives and how they really are.

Accordingly, the productive critique does not prevent thinking against DP. On the contrary, showing the onto-epistemological discontinuities of DP with pre-digital psychiatry enables the possibility of producing well-founded comparisons and critiques. When thinking with and against DP, it is possible to show that what the app users see in the self-tracking charts are the dynamic interactions between their behaviors and emotions, with broad statistical parameters and neuropsychological assumptions. The high velocity, ubiquity, and plasticity of human–computer’s interactions overlap both scales, producing the effect of personalized truth. In other words, my depression overlaps with the generic diagnostic category, and the minimal gaps between them guide me on how to manage my mental health. Thus, thinking critically with and against DP involves noting that what has been personalized are not the neuropsychological categories through which one knows oneself but how we relate to them: DP provides personalized pathways to normalization and neuropsychologization.

Conclusion

In this article, we make a critical analysis of the emergence of DP in psychiatry. Instead of conceiving DP as a new tool to get to know old objects—or solve old neuropsychiatric problems—we underline the potential productive effects of this new onto-epistemological matrix that reinvents both psychiatric objects and subjectivities.

First, we show that DP's scientific prestige and cultural diffusion are directly linked with its promise to solve some of the major impasses left open by the neuro-turn in psychiatry—particularly, the RDoC project. New digital technologies and AI burst onto the psychiatric scene carrying the promise of connecting clinical phenotypes with hard biological evidence and producing a new sort of neuropsychiatric knowledge that relied neither on experts nor patients’ subjectivities. Nevertheless, we show that rather than finding new psychobiological brain-mind mediators, DP is embedded in what we have called the “neuro-digital complex,” which redefines some important features of pre-digital psychiatry. Analyzed as a new onto-epistemic assembly, this complex is a hybrid assembly of digital technologies embedded in our daily lives, discourses, and practices that redefine the brain in terms of emergent patterns, and of the mind, in terms of unintended human–computer micro-interactions. Within this new framework, the mind and the brain are conceived as computable data which can be analyzed with AI, allowing DP to produce/identify new patterns of mental health by combining clinical and biological evidence.

Secondly, the article analyzes that DP's central promise is to produce personalized diagnostics and tailored interventions, reducing the elapsed time between diagnosis, treatment, and/or prevention-orientated actions. Yet, to make sense of intra-individual information, DP must compare it with population statistics, grounded truth data sets built with pre-digital neuropsychological assumptions, and human categorization processes. Thus, digital phenotypes are not neutral digital mirrors for self-knowledge because they encode different forms of cultural, scientific, and technological biases. Even so, this fact contrasts with the effect of reified and personalized neuropsychological truth they produce. Both experts and app users have the impression that DP gives them access to individuals’ neuropsychological core. Instead of analyzing this gap between the “objective digital truth” and “the subjective effect of personalized truth” as an ideological fact, we have analyzed the sociotechnical condition of possibilities for the emergence of DP's new regime of truth.

Thirdly, we show that the DP regime of truth is the effect of an assembly of pre-digital neuropsychological cultures and categories, self-tracking cultures and subjectivities, and the new possibilities opened by technological devices embedded in our everyday lives. From this viewpoint, the gap between how DP produces the truth about us and how we perceive it becomes productive: it measures the distance between whom we think we are and who we actually are. Therefore, we conclude that even though DP still participates in pre-digital processes of neuropsychologyzation, it does so in a new way. Rather than providing personalized diagnoses and treatments—as DP's techno-utopia claims—what has been individualized are the pathways to normalization and neuropsychologization.

Finally, it is important to emphasize that the scope of this article should be considered within certain limits. Its theoretical approach to DP leaves opens the question of concrete practices. Questions such as “how do app users and experts interpret the personalized neuro-digital truth?”; “What kind of other cultural and scientific biases is DP eventually reproducing?”; or “What distinguishes DP personalized truth from the one produced by other self-tracking cultures?” remain unresolved in this article. Beyond these limitations, we hope this work will contribute to future critical analyses of DP and open new questions regarding the specificities of the neuro-digital complex.

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Fondo Nacional de Desarrollo Científico y Tecnológico (grant number FONDECYT POSTDOCTORADO/2020-3200944, Millennium Science Initiative Program, grant ICS13, Millennium Science Initiative Program, grant NCS20). ÁJ-M received funding from ANID/Millennium Science Initiative Program, grant NCS2021_81 and ICS13_005, and ANID/FONDECYT POSTDOCTORADO/2020-3200944.

ORCID iDs

Rodrigo De La Fabián

Álvaro Jiménez-Molina

Francisco Pizarro Obaid

Notes

References

American Psychiatric Association (APA). (2013) Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5). Washington, DC: American Psychiatric Association.

Balaskas

Schueller

Cox

, et al. (2021) Ecological momentary interventions for mental health: A scoping review. PLoS ONE 16(3): e0248152.

Barad

(2003) Posthumanist performativity: Toward an understanding of how matter comes to matter. Signs: Journal of Women in Culture and Society 28(3): 801–831.

Barad

(2007) Meeting the Universe Halfway: Quantum Physics and the Entanglement of Matter and Meaning. Durham: Duke University Press.

Baumeister

Montag

(2019) Digital Phenotyping and Mobile Sensing: New Developments in Psychoinformatics. Switzerland: Springer Nature.

Baumgartner

(2021) Precision medicine and digital phenotyping: Digital medicine’s way from more data to better health. Big Data & Society 8(2): 1–12.

Bemme

Brenman Semel

(2020) The subjects of digital psychiatry. Somatosphere. 13 October, n.p. Available at: http://somatosphere.net/2020/subjects-of-digital-psychiatry.html/ (accessed 29 November 2022).

Benjamin

(2019) Race After Technology: Abolitionist Tools for the New Jim Code. Cambridge: Polity Press.

Benoit

Onyeaka

Keshavan

, et al. (2020) Systematic review of digital phenotyping and machine learning in psychosis spectrum illnesses. Harvard Review of Psychiatry 28(5): 296–304.

10.

Berríos

(2002) The History of Mental Symptoms: Descriptive Psychopathology Since the Nineteenth Century. Cambridge: Cambridge University Press.

11.

Bidargaddi

Musiat

Makinen

, et al. (2016) Digital footprints: Facilitating large-scale environmental psychiatric research in naturalistic settings through data from everyday technologies. Molecular Psychiatry 22(2): 164–169.

12.

Birk

Lavis

Lucivero

, et al. (2021) For what it’s worth. Unearthing the values embedded in digital phenotyping for mental health. Big Data & Society 8(2): 1–5.

13.

Birk

Samuel

(2020) Can digital data diagnose mental health problems? A sociological exploration of ‘digital phenotyping’. Sociology of Health & Illness 42(8): 1873–1887.

14.

Bradstreet

Allan

Gumley

(2019) Adverse event monitoring in mHealth for psychosis interventions provides an important opportunity for learning. Journal of Mental Health 28(5): 461–466.

15.

Browne

(2015) Dark Matters: On the Surveillance of Blackness. Durham: Duke University Press.

16.

Bucher

(2018) If … Then: Algorithmic Power and Politics. New York: Oxford University Press.

17.

Camacho

Brady Junior

Lizano

, et al. (2021) Advancing translational research through the interface of digital phenotyping and neuroimaging: A narrative review. Biomarkers in Neuropsychiatry 4: 1–8.

18.

Cobb

(2020) The Idea of the Brain: A History. New York: Basic Books.

19.

Csikszentmihalyi

Larson

(1987) Validity and reliability of the experience-sampling method. Journal of Nervous and Mental Disease 175(9): 526–536.

20.

Currey

Torous

(2022) Digital phenotyping correlations in larger mental health samples: Analysis and replication. BJPsych Open 8(e106): 1–7.

21.

Cuthbert

(2014) The RDoC framework: Facilitating transition from ICD/DSM to dimensional approaches that integrate neuroscience and psychopathology. World Psychiatry 13(1): 28–35.

22.

Cuthbert

Insel

(2013) Toward the future of psychiatric diagnosis: The seven pillars of RDoC. BMC Medicine 11(1): 126.

23.

Danziger

(1997) Naming the Mind. How Psychology Found its Language. London: SAGE.

24.

Dawkins

(1983) The Extended Phenotype: The Long Reach of the Gene. Oxford: Oxford University Press.

25.

Demazeux

(2019) L’éclipse du symptôme. L’observation clinique en psychiatrie 1800–1950. Paris: Ithaque.

26.

Dennett

(1991) Consciousness Explained. Boston: Little Brown.

27.

De Vos

(2013) Psychologization and the Subject of Late Modernity. New York: Palgrave Macmillan.

28.

De Vos

(2020) Digital therapeutic culture. In: Nehring D, Madsen OJ, Cabanas et al. (eds) The Routledge International Handbook of Global Therapeutic Cultures. London: Routledge, pp.233–244.

29.

Doryab

Villalba

Chikersal

, et al. (2019) Identifying behavioral phenotypes of loneliness and social isolation with passive sensing: Statistical analysis, data mining and machine learning of smartphone and fitbit data. JMIR mHealth and UHealth 7(7): e13209.

30.

Dourish

Gómez Cruz

(2018) Datafication and data fiction: Narrating data and narrating with data. Big Data & Society 5(2): 1–10.

31.

Faurholt-Jepsen

Mads

Christensen

, et al. (2020) The effect of smartphone-based monitoring on illness activity in bipolar disorder: The MONARCA II randomized controlled single-blinded trial. Psychological Medicine 50(5): 838–848.

32.

Finn

(2017) What Algorithms Want: Imagination in the Age of Computing. Cambridge: MIT Press.

33.

Foucault

(1972) Archaeology of Knowledge and Discourse on Language. New York: Pantheon Books.

34.

Foucault

(1976) La fonction politique de l’intellectuel. In: Foucault

(ed) (2001) Dits et Écrits II. Paris: Gallimard, pp.109–114.

35.

Foucault

(1982) Technologies of the self. In: Rabinow

(ed) (1997) Michel Foucault. Ethics, Subjectivity and Truth. The Essential Works of Michel Foucault, 1954–1984, Volume 1. New York: The New Press, pp.223–252.

36.

Foucault

(2008) The Birth of Biopolitics. New York: Palgrave Macmillan.

37.

Gillespie

(2014) The relevance of algorithms. In: Gillespie

Boczkowski

Kirsten

(eds) Media Technologies: Essays on Communication, Materiality, and Society. Cambridge, MA: The MIT Press, pp.167–193.

38.

Ginger. (n.d.) Ginger. Available at: https://www.ginger.com/ (accessed 29 November 2022).

39.

Ginger.io. (2015) Ginger.io. 18 February. Available at: https://www.youtube.com/watch?v=l4UqST9zuoo (accessed 29 November 2022).

40.

Green

Svendsen M

(2021) Digital phenotyping and data inheritance. Big Data & Society 8(2): 1–5.

41.

Grosman

Reigeluth

(2019) Perspectives on algorithmic normativities: Engineers, objects, activities. Big Data & Society 6(2): 1–12.

42.

Haag

(2007) Biomarkers trump behavior in mental illness diagnosis. Nature Medicine 13(1): 3.

43.

Hayles

(1999) How We Became Posthuman: Virtual Bodies in Cybernetics, Literature, and Informatics. Chicago: The University of Chicago Press.

44.

Henson

Barnett

Keshavan

, et al. (2020) Towards clinically actionable digital phenotyping targets in schizophrenia. Npj Schizophrenia 6(13): 1–7.

45.

Hesse

(2008) ‘Bytes of Life’. The Washington Post, 9 September, pp. 1–5. Available at: https://www.washingtonpostcom/wp-dyn/content/article/2008/09/08/AR2008090802681.html (accessed 29 November 2022).

46.

Hidalgo-Mazzei

Young

Vieta

, et al. (2018) Behavioural biomarkers and mobile mental health: A new paradigm. International Journal of Bipolar Disorders 6(9): 1–4.

47.

Huckvale

Venkatesh

Christensen

(2019) Toward clinical digital phenotyping: A timely opportunity to consider purpose, quality, and safety. Npj Digital Medicine 2(1): 1–11.

48.

Huys

Browning

Paulus

, et al. (2021) Advances in the computational understanding of mental illness. Neuropsychopharmacology 46(1): 3–19.

49.

Insel

(2010) Research domain criteria (RDoC): Toward a new classification framework for research on mental disorders. American Journal of Psychiatry 167(7): 748–751.

50.

Insel

(2013) ‘Psychiatry’s Guide is Out of Touch with Science, Experts Say’. Interviewed by Pam Belluck and Benedict Carey, The New York Times, 6 May. Available at: https://www.nytimes.com/2013/05/07/health/psychiatrys-new-guide-falls-short-experts-say.html (accessed 29 November 2022).

51.

Insel

(2018) Digital phenotyping: A global tool for psychiatry. World Psychiatry 17(3): 276–277.

52.

Insel

(2017a) ‘Star Neuroscientist Tom Insel Leaves the Google-Spawned Verily for … a Startup?’ Interviewed by: Adam Rogers, Wired, 11 May. Available at: https://www.wired.com/2017/05/star-neuroscientist-tom-insel-leaves-google-spawned-verily-startup/ (accessed 29 November 2022).

53.

Insel

(2017b) ‘The Smartphone Psychiatrist’. Interviewed by: David Dobbs, The Atlantic, July/August Available at: https://www.theatlantic.com/magazine/archive/2017/07/the-smartphone-psychiatrist/528726/ (29 November 2022).

54.

Insel

(2017c) Digital phenotyping: Technology for a new science of behavior. JAMA 318(13): 1215–1216.

55.

Jacobson

Weingarden

Wilhelm

(2019) Using digital phenotyping to accurately detect depression severity. The Journal of Nervous and Mental Disease 207(10): 893–896.

56.

Jain

Powers

Hawkins

, et al. (2015) The digital phenotype. Nature Biotechnology 33(5): 462–463.

57.

Kapur S, Phillips AG, and Insel TR (2012) Why has it taken so long for biological psychiatry to develop clinical tests and what to do about it? Molecular Psychiatry 17: 1174–1179.

58.

Kleiman EM, Nock MK, Turner BJ, et al. (2018) Digital phenotyping of suicidal thoughts. Depression and Anxiety 35(7): 601–608.

59.

Kozak

Cuthbert

(2016) The NIMH research domain criteria initiative: Background, issues, and pragmatics. Psychophysiology 53(3): 286–297.

60.

Kupfer

(2013) “Experts say ‘Psychiatry’s Guide is out of Touch with Science.’ What Science?” Interviewed by Jabob Sullum, Reason, 5 July. Available at: https://reason.com/2013/05/07/experts-say-psychiatrys-guide-is-out-of/ (accessed 29 November 2022).

61.

Latour

(2004) Why has critique run out of steam? From matters of fact to matters of concern. Critical Inquiry 30(2): 225–248.

62.

Latour

(2006) Avons-nous besoin de ‘paradigmes’? In: Latour

(ed) Chroniques D’un Amateur de Sciences. Paris: Mines Paris, Les Presses, pp.27–30.

63.

Ledford

(2015) Director of US Mental-Health Institute leaves for Google. Nature News, 15 September.

64.

Liang

Zheng

Zeng

(2019) A survey on big data-driven digital phenotyping of mental health. Information Fusion 52: 290–307.

65.

Littlefield

Johnson

(2012) The Neuroscientific Turn: Transdisciplinarity in the Age of the Brain. Ann Arbor: University of Michigan Press.

66.

Loi

(2019) The digital phenotype: A philosophical and ethical exploration. Philosophy & Technology 32(1): 155–171.

67.

Lucivero

Hallowell

(2021) Digital/computational phenotyping: What are the differences in the science and the ethics? Big Data & Society 8(2): 1–5.

68.

Lupton

(2014) Self-Tracking cultures: Towards a sociology of personal informatics. In: 26^th Australian Computer–Human Interaction Conference on Designing Futures: The Future of Designing, Sydney, New South Wales, Australia, 2–5 December 2014, pp. 77–86.

69.

Lupton

(2016) The Quantified Self: A Sociology of Self-Tracking. Cambridge: Polity.

70.

Markowetz

Błaszkiewicz

Montag

, et al. (2014) Psycho-informatics: Big data shaping modern psychometrics. Medical Hypotheses 82(4): 405–411.

71.

Martin-Key

Schei

Barker

, et al. (2021) The current state and diagnostic accuracy of digital mental health assessment tools for psychiatric disorders: Protocol for a systematic review and meta-analysis. JMIR 10(1): e25382.

72.

Mayer-Schönberger

Cukier

(2017) Big Data: A Revolution that will Transform How We Live, Work and Think. London: John Murray, 2017.

73.

Melcher

Hays

Torous

(2020) Digital phenotyping for mental health of college students: A clinical review. Evidence Based Mental Health 23(4): 161–166.

74.

Mendes

Moura

Van de Ven

, et al. (2022) Sensing apps and public data sets for digital phenotyping of mental health: Systematic review. Journal of Medical Internet Research 24(2): e28735.

75.

Mindstong (n.d.a) Mindstrong. Mental health maid for you. Available at: https://mindstrong.com/ (accessed 30 August 2022).

76.

Mindstrong (n.d.b) Mindstrong. Using science to help us seek the truth. Available at: https://mindstrong.com/science/ (accessed 30 August 2022).

77.

Minstrong (n.d.bc) Mindstrong.Your resources as a Mindstrong member. Available at: https://mindstrong.com/our-services/ (accessed 30 August 2022).

78.

Mohr

Shilton

Hotopf

(2020) Digital phenotyping, behavioral sensing, or personal sensing: Names and transparency in the digital age. Npj Digital Medicine 3(45): 1–2.

79.

Mohr

Zhang

Schueller

(2017) Personal sensing: Understanding mental health using ubiquitous sensors and machine learning. Annual Review of Clinical Psychology 13(8): 11.1–11.25.

80.

Nakamura

(2008) Digitizing Race: Visual Cultures of the Internet. Minneapolis: University of Minnesota Press.

81.

Nehring

Madsen

Cabanas

, et al. (2020) Introduction. Therapeutic global cultures from a multidisciplinary perspective: present and future challenges. In: Nehring D, Madsen OJ, Cabanas E, et al. (eds) The Routledge International Handbook of Global Therapeutic Cultures. London: Routledge, pp. 3–13.

82.

Onnela

(2021) Opportunities and challenges in the collection and analysis of digital phenotyping data. Neuropsychopharmacology 46(1): 45–54.

83.

Onnela

(n.d.) Digital phenotyping and Beiwe Research Platform. Onnela Lab. Harvard T.H. Chan School of Public Health. Available at: https://www.hsph.harvard.edu/onnela-lab/beiwe-research-platform/ (29 November 2022).

84.

Onnela

Rauch S

(2016) Harnessing smartphone-based digital phenotyping to enhance behavioral and mental health. Neuropsychopharmacology 41(7): 1691–1696.

85.

Patrick

Hajcak

(2016) RDoc: Translating promise into progress. Psychophysiology 53(3): 415–424.

86.

Peterson

(2020) Editorial: Biomarkers in precision medicine for mental illnesses. Journal of Child Psychology and Psychiatry and Allied Disciplines 61(12): 1279–1281.

87.

Piau

Wild

Mattek

, et al. (2019) Current state of digital biomarker technologies for real-life, home-based monitoring of cognitive function for mild cognitive impairment to mild Alzheimer disease and implications for clinical care: Systematic review. Journal of Medical Internet Research 21(8): e12785.

88.

Pickersgill

(2019) Digitising psychiatry? Sociotechnical expectations, performative nominalism and biomedical virtue in (digital) psychiatric praxis. Sociology of Health & Illness 41: 16–30.

89.

Proudfoot

Clarke

Birch

, et al. (2013) Impact of a mobile phone and web program on symptom and functional outcomes for people with mild-to-moderate depression, anxiety and stress: A randomised controlled trial. BMC Psychiatry 13(1): 1–12.

90.

Rhim

Lee

Han

(2020) Tracking and modeling subjective well-being using smartphone-based digital phenotype. In: UMAP’20: Proceedings of the 28th ACM Conference on User Modeling, Adaptation and Personalization, Genoa, Italy, 14–17 July 2020, pp. 211–220.

91.

Richards

Lillicrap

Beaudoin

, et al. (2019) A deep learning framework for neuroscience. Nature Neuroscience 22(11): 1761–1770.

92.

Rohani

Faurholt-Jepsen

Kessing

, et al. (2018) Correlations between objective behavioral features collected from mobile and wearable devices and depressive mood symptoms in patients with affective disorders: Systematic review. JMIR Mhealth and Uhealth 6(8): 1–21.

93.

Rose

(1998) Inventing Ourselves: Psychology, Power, and Personhood. Cambridge: University Press.

94.

Rose

(2003) Neurochemical selves. Society 41(1): 46–59.

95.

Rose

(2019) Our Psychiatric Future: The Politics of Mental Health. Cambridge: Polity Press.

96.

Rose

Abi-Rached

(2013) Neuro: The New Brain Sciences and the Management of the Mind. Princeton: Princeton University Press.

97.

Rouvroy

Berns

(2013) Gouvernementalité algorithmique et perspectives d’émancipation Le disparate comme condition d’individuation par la relation ? Réseaux 177: 163–196.

98.

Rowe

(2021) Social determinants of health in the big data mode of population health risk calculation. Big Data & Society 8(2): 1–12.

99.

Saccaro

Amatori

Cappelli

, et al. (2021) Portable technologies for digital phenotyping of bipolar disorder: A systematic review. Journal of Affective Disorders 295: 323–338.

100.

Schork

(2015) Personalized medicine: Time for one-person trials. Nature 520: 609–611.

101.

Semel

(2021) Listening like a computer: Attentional tensions and mechanized care in psychiatric digital phenotyping. Science, Technology & Human Values 47(2): 266–290.

102.

Spinazze

Rykov

Bottle

, et al. (2019) Digital phenotyping for assessment and prediction of mental health outcomes: A scoping review protocol. BMJ Open 9(12): 1–6.

103.

Spitzer

Endicott

(1968) DIAGNO: A computer program for psychiatric diagnosis utilizing the differential diagnostic procedure. Archives of General Psychiatry 18(6): 746–756.

104.

Stanghellini

Leoni

(2020) Digital phenotyping: Ethical issues, opportunities, and threats. Frontiers in Psychiatry 11: 73.

105.

Stephens-Davidowitz

(2017) Everybody Lies: Big Data, New Data, and What the Internet can Tell us About Who We Really Are. New York: Harper Colllins.

106.

Stone

Shiffman

(1994) Ecological momentary assessment (EMA) in behavioral medicine. Annals of Behavioral Medicine 16(3): 199–202.

107.

Taschereau-Dumouchel

Michel

Lau

, et al. (2022) Putting the “mental” back in “mental disorders”: A perspective from research on fear and anxiety. Molecular Psychiatry 27(3): 1322–1330.

108.

Tekin

(2020) Is big data the new stethoscope? Perils of digital phenotyping to address mental illness. Philosophy & Technology 34: 447–461.

109.

Terranova

(2004) Network Culture: Politics for the Information Age. London: Pluto Press.

110.

Torous

Bucci

Bell

, et al. (2021) The growing field of digital psychiatry: Current evidence and the future of apps, social media, chatbots, and virtual reality. World Psychiatry 20(3): 318–335.

111.

Torous

Kiang

Lorme

, et al. (2016) New tools for new research in psychiatry: A scalable and customizable platform to empower data driven smartphone research. JMIR Mental Health 3(2): e16.

112.

Torous

Onnela

Keshavan

(2017) New dimensions and new tools to realize the potential of RDoC: Digital phenotyping via smartphones and connected devices. Translational Psychiatry 7: e1053.

113.

Turkheimer

Hellyer

Kehagia

, et al. (2019) Conflicting emergences. Weak vs. strong emergence for the modelling of brain function. Neuroscience and Biobehavioral Reviews 99: 3–10.

114.

Vahabzadeh

Sahin

Kalali

(2016) Digital suicide prevention: Can technology become a game changer? Innovations in Clinical Neuroscience 13(5–6): 16–20.

115.

Vegter

Zwart

HAE

Van Gool

(2021) The funhouse mirror: The I in personalised healthcare. Life Sciences, Society and Policy 17(1): 1–15.

116.

Vidal

Ortega

(2017) Being Brains: Making the Cerebral Subject. New York: Fordham University Press.

117.

Vilar

Pérez-Sola

Blasco

, et al. (2019) Translational research in psychiatry: The research domain criteria project (RDoC). Revista de Psiquiatría y Salud Mental (English Edition) 12(3): 187–195.

118.

Wang

Miller

(2019) Just-in-the-moment adaptive interventions (JITAI): A meta-analytical review. Health Communication 35(12): 1531–1544.

119.

Wang

Nikola

Creighton

, et al. (2021) HOPES: An integrative digital phenotyping platform for data collection, monitoring, and machine learning. Journal of Medical Internet Research 23(3): e23984.

120.

World Health Organization (WHO) (2018) mHealth. Use of appropriate digital technologies for public health. In: Seventy-First World Health Assembly. Provisional agenda item 12.4, A71/20. Available at: https://apps.who.int/gb/ebwha/pdf_files/WHA71/A71_20-en.pdf (accessed 29 November 2022).

121.

Yarkoni

Westfall

(2017) Choosing prediction over explanation in psychology: Lessons from machine learning. Perspectives on Psychological Science 12(6): 1100–1122.

122.

Yuste

(2015) From the neuron doctrine to neural networks. Nature Reviews Neuroscience 16(8): 487–497.

123.

Zulueta

Piscitello

Rasic

, et al. (2018) Predicting mood disturbance severity with mobile phone keystroke metadata: A BiAffect DP study. Journal of Medical Internet Research 20(7): 1–10.

A critical analysis of digital phenotyping and the neuro-digital complex in psychiatry

Abstract

Keywords

Introduction

Neuro-turn impasses and the background of digital phenotyping

The neuro-digital complex: The digitalization of the mind and the de-substantialization of the brain

“Closing the loop”: The promise of digital phenotyping mental health interventions

Thinking with and against digital phenotyping

Conclusion

Footnotes

Declaration of conflicting interests

Funding

ORCID iDs

Notes

References