2.4 A Behavioural Science of the Individual?

The dictum “first analyze, then aggregate” (by Peter Molenaar as quoted in Rose, 2016), advocates an order reversal of the methods commonly employed by the behavioural sciences to advance scientific knowledge: Measurement outcomes of psychological variables observed in samples of individuals that share some identity are first aggregated, then analyzed, in order to decide whether the statistical properties of the variables realized in the sample can be considered general regularities or true characteristics of the identity under scrutiny. As such, nomothetic behavioural science produces knowledge about psychological and behavioural phenomena that are expected to be robust at the level of aggregated wholes. However, contrary to the applied branches of economic, sociological and political science, the scientist practitioner generally does not deal with the behaviour of aggregated wholes such as economies, societies and groups of voters. The domain in which they have to apply this knowledge concerns the explanation, intervention, or prediction of the behaviour of particular patients, children, or employees, that is, they deal with the individuals that constitute the aggregated wholes. As mentioned earlier, recent studies have shown that applying the statistical syllogism to nomothetic knowledge in psychology may be very problematic.

One could argue that the rules of sample-based statistical inference do not warrant generalizations to a particular case, so one should not engage in such inductive reasoning in the first place. This is of course an unsatisfactory solution and about 15 years ago Molenaar presented an alternate path for psychological science in a paper entitled: “A Manifesto on Psychology as Idiographic Science: Bringing the Person Back Into Scientific Psychology, This Time Forever” (Molenaar, 2004). It’s still too early to call whether it will indeed be forever, but the past decade did see a surge of scientific studies employing a person-centered or idiographic approach, most prominently in the field of psychopathology. Many researchers have been exploring what can be described as a small data paradigm (cf. Hekler et al., 2019), which departs from the same goals and principles as so-called precision, or personalized medicine (I. S. Chan & Ginsburg, 2011), promising to yield new insights and better tools for scientist practitioners to apply to the particular cases they encounter in their daily practice (David, Marshall, Evanovich, & Mumma, 2018; see e.g., Schiepek et al., 2016). In general, the approach entails studying a limited number of cases intensively, for example, by collecting densely sampled multivariate time series data that represent self-reports of emotional states or well-being, using the so-called Experience Sampling Method, or Ecological Momentary Assessment (Conner, Tennen, Fleeson, & Barrett, 2009; Schiepek, Stöger-Schmidinger, Aichhorn, Schöller, & Aas, 2016; Wichers, Groot, & Psychosystems, 2016). Naturally, this change of focus in data collection from static sampling from homogeneous populations, to dynamic process monitoring in individuals is accompanied by the development of new theories, models and analytic techniques, of which the network approach to psychopathology has probably been the most popular and revolutionary approach (cf. McNally, 2019; Robinaugh, Hoekstra, Toner, & Borsboom, 2020).

There appears to be a substantial gap between the conceptual changes in the theoretical ideas about the nature of the system in which psychological phenomena can be observed and the analytic techniques and rules of inference employed to study such systems. Consider the following description of the personalized approach:

“The personalized approach to psychopathology conceptualizes mental disorder as a complex system of contextualized dynamic processes that is non-trivially specific to each individual, and seeks to develop formal idiographic statistical models to represent these individual processes.” (Wright & Woods, 2020)

Although many studies using the personalized approach indeed depart from the idea that human behaviour, whether pathological or not, should be considered to arise from a complex dynamical system (see e.g., Cramer et al., 2016), few studies actually make use of the methods and analyses from Complexity Science that were developed to study such systems. It appears to be the case that in terms of methods and analytic techniques, the idiographic revolution is taking place right now, but the complexity revolution has yet to occur. Two recent reviews of the personalized approach (Piccirillo & Rodebaugh, 2019; Wright & Woods, 2020) fail to discusses, or even mention, the use of complexity methods to study multivariate time series data in the context of psychopathology. Such studies do exist and have been a part of the scientific record for at least as long as studies using the Gaussian Graphical Model to model multivariate ESM data (Delignières, Fortes, Ninot, et al., 2004; see e.g., Guastello, Koopmans, & Pincus, 2008; Lichtwarck-Aschoff, Hasselman, Cox, Pepler, & Granic, 2012; Olthof, Hasselman, Strunk, Aas, et al., 2019; Olthof, Hasselman, Strunk, Rooij, et al., 2019; Schiepek, 2003, 2009; Van Geert & Fischer, 2009).

It is the purpose of this book to bridge the gap and introduce complexity methods that will be essential tools for the analytic toolbox of a behavioural science of the individual.