Formal representation of ambulatory assessment protocols in HTML5 for human readability and computer execution

Behavior Research Methods (2019) 51:2761–2776 https://doi.org/10.3758/s13428-018-1148-y Formal representation of ambulatory assessment protocols in HTML5 for human readability and computer execution Nikolaos Batalas1 · Vassilis-Javed Khan1 · Minita Franzen2 · Panos Markopoulos1 · Marije aan het Rot2 Published online: 7 November 2018 © The Author(s) 2018 Abstract Ambulatory assessment (AA) is a research method that aims to collect longitudinal biopsychosocial data in groups of individuals. AA studies are commonly conducted via mobile devices such as smartphones. Researchers tend to communicate their AA protocols to the community in natural language by describing step-by-step procedures operating on a set of materials. However, natural language requires effort to transcribe onto and from the software systems used for data collection, and may be ambiguous, thereby making it harder to reproduce a study. Though AA protocols may also be written as code in a programming language, most programming languages are not easily read by most researchers. Thus, the quality of scientific discourse on AA stands to gain from protocol descriptions that are easy to read, yet remain formal and readily executable by computers. This paper makes the case for using the HyperText Markup Language (HTML) to achieve this. While HTML can suitably describe AA materials, it cannot describe AA procedures. To resolve this, and taking away lessons from previous efforts with protocol implementations in a system called TEMPEST, we offer a set of custom HTML5 elements that help treat HTML documents as executable programs that can both render AA materials, and effect AA procedures on computational platforms. Keywords Ambulatory assessment · Experience sampling · HTML · Protocol representations · Data collection software systems Introduction Scientists in the social and medical sciences use intensive longitudinal methods (ILM) (Bolger & Laurenceau, 2013) for the repetitive sampling of individuals in the context of their daily lives and routines over extended periods of time. ILM are preferable over retrospective reports because the results are shown to be more valid for measuring actual experience, not having been compromised by memory biases (Moskowitz et al., 2009). Ambulatory assessment (AA) (Fahrenberg et al., 2007) is a type of ILM that employs several techniques, including the experience sampling method (ESM) (Csikszentmihalyi & Larson, 2014), the ecological momentary assessment method (EMA) (Stone & Shiffman, 1994), as well as monitoring environmental and  Nikolaos Batalas 1 Department of Industrial Design, Eindhoven University of Technology, Eindhoven, The Netherlands 2 Department of Psychology, University of Groningen, Groningen, The Netherlands physiological parameters through the use of electronic sensor devices, such as GPS, ambient light and noise sensors, or heart-rate monitors. The utilization of such technological elements makes smartphones very suitable for executing AA studies. As such, researchers conduct AA studies through software systems that help them implement and manage the study’s protocol. A protocol is the plan for collecting data (Vogt & Johnson, 2011). In the context of a specific method, such as AA, the protocol includes two components: (1) detailed definitions of the materials (e.g., the instruments to be used), and (2) instructions on how the data collection procedures are to take place (e.g., detailed descriptions of how the instruments are deployed). There is agreement within the scientific community that the detailed representation of study protocols is important for research to be effective, replicable, and implementable (Michie et al., 2011), and calls are being made to share not only data but also materials and code (Nosek et al., 2015). Yet, AA researchers emphasize the persistence of several issues in successfully reporting research that captures momentary assessments, contributing to a broader ‘replication crisis’ (or reproducibility crisis) in the social and 2762 medical sciences (Stone, 2017). Among those are issues of reporting on the data acquisition interface, and on the details of the sampling process (Stone & Shiffman, 2002), i.e., the materials and procedures of an AA protocol. As AA methods have become increasingly prevalent and accepted in the social and medical sciences over the last several decades (Riese, 2017), including broadly in the field of psychology, addressing the need to share AA protocols in a way that makes them easy to understand and implement at the same time is a noteworthy undertaking. To describe AA protocols, researchers most often use natural language (as in Gunthert et al., 2007), and seldom other high-level representations, such as flowcharts (as in Ellis-Davies et al., 2012). In the present paper, we discuss why these practices make describing digital materials and procedures cumbersome, and their replication prone to errors. Subsequently, we examine the software tools available to researchers for the implementation of AA studies, and show that they do not cater to the need for representing AA protocols to third parties in sufficient detail. To address this gap, we put forward a set of requirements for adequate AA protocol representations, which have to do with being both human readable and computer executable. Finally, we propose a solution to satisfy the proposed requirements, one that takes advantage of the ubiquitous Hypertext Markup Language (HTML) and the Web browser. Importantly, while HTML documents are easy to read by both humans and machines, they cannot represent executable processes. Thus, we contribute a set of HTML5 custom elements that help treat an HTML document as an executable program. This document is representative of an AA protocol in both its materials and procedures, and serves as the single source for both the description and the implementation of an AA study. Using HTML documents is thus considered to reduce errors, and may foster easier replication of AA studies even when researchers may still use different systems to carry them out. Ambulatory assessment As “a class of methods that makes use of mobile technology to understand people’s biopsychosocial processes in natural settings, in real-time, and on repeated occasions” (Conner & Mehl, 2015), AA methods share the motivations and include the procedures of ESM and EMA. In the remainder of this paper, as AA studies tend to collect both survey and sensor data, we will use the term AA as an umbrella term for various examples of smartphone-based ILM, including ESM and EMA. ESM, where participants are asked to regularly report their subjective experience, has traditionally focused only on survey data. Although EMA, where the interest of researchers extends also to physiological processes, has included sensor data, these have generally Behav Res (2019) 51:2761–2776 not been collected continuously. Examples of applications of AA methods can be found (...truncated)