Assessing causal claims about complex engineered systems with quantitative data: internal, external, and construct validity

Abstract

Engineers seek to design systems that will produce an intended change in the state of the world. How are we to know if a system will behave as intended? This article addresses ways that this question can be answered. Specifically, we focus on three types of research validity: (1) internal validity, or whether an observed association between two variables can be attributed to a causal link between them; (2) external validity, or whether a causal link generalizes across contexts; and (3) construct validity, or whether a specific set of metrics corresponds to what they are intended to measure. In each case, we discuss techniques that may be used to establish the corresponding type of validity: namely, quasi‐experimental design, replication, and establishment of convergent‐discriminant validity and reliability. These techniques typically require access to data, which has historically been limited for research on complex engineered systems. This is likely to change in the era of “big data.” Thus, we discuss the continued utility of these validity concepts in the face of advances in machine learning and big data as they pertain to complex engineered sociotechnical systems. Next, we discuss relationships between these validity concepts and other prominent approaches to evaluating research in the field. Finally, we propose a set of criteria by which one may evaluate research utilizing quantitative observation to test causal theory in the field of complex engineered systems.

Psychological foundations of explainability and interpretability in artificial intelligence

“…10 Explainability and interpretability are both "ilities" and exhibit similar difficulties associated with their measurement. "Ilities" have historically been subject to problems of both polysemy, meaning that the same terms are frequently used to describe distinct concepts, and synonymy, meaning that different terms sometimes refer to the same underlying construct [28]. Furthermore, these terms entail a significant social component that cannot be disentangled from core values of users, designers, and decision-makers.…”

Communicating Meaning in the Intelligence Enterprise

“…These five insights are based upon a large body of empirically validated theory, FTT, which has demonstrated applicability across several contexts. Beyond laboratory studies (see Reyna, 2012a, for a review), FTT has demonstrated the ability to make actionable predictions in the domains of intelligence analysis (Reyna et al, 2014), medical decision making (Reyna, 2008), legal reasoning (Brainerd, Reyna, & Poole, 2000; Reyna, Mills, Estrada, & Brainerd, 2006), advertising (LaTour, LaTour, & Brainerd, 2014), public health (Reyna & Mills, 2014), text comprehension (Reyna, Corbin, Weldon, & Brainerd, 2016; Reyna & Kiernan, 1995), engineering (Broniatowski, 2018; Broniatowski & Tucker, 2017), and others. Indeed, the wide reach of FTT’s findings suggests that they are broadly applicable across the intelligence enterprise, and especially at the interface between intelligence agencies and policy makers.…”

Building the tower without climbing it: Progress in engineering systems

“…In general, approaches that call for replacing requirements with optimization may prove to be counterproductive because they do not incorporate decision makers’ expertise. As I have argued elsewhere, requirements typically establish the gist of “success” or “failure” for a given design engineer; however, novices and experts may interpret these requirements differently. Specifically, experts possess the contextual knowledge to know when “bending” the requirements can be helpful, whereas novices might use a more rote strategy, treating requirements as strict thresholds.…”

“…This put in place a process that continues to reshape the way systems engineering faculty think about, and evaluate, work in the field. For example, Broniatowski and Tucker proposed that ES research would benefit from adopting validity concepts that are widespread in the social sciences. Specifically, competing theories can be assessed for their internal validity (whether a proposed cause leads to a proposed effect), external validity (whether it generalizes across contexts), and construct validity (whether its underlying theoretical constructs have been demonstrated to exist) .…”