Cognitive Complexity Lab

If you are looking for Automated Integrative Complexity, go to

What do we study?

In the cognitive complexity lab, we study the causes and consequences of communication complexity.  For example, we are at the forefront of research for both the natural language processing of complexity and applying complexity research to national security (click on the links to learn more).

Along the way, we have studied a wide array of persons: College students, U.S. Presidents and Presidential Candidates, terrorists, Canadian Prime Ministers, famous religious converts, famous atheists, and famous liars such as Richard Nixon. We have also studied the complexity effects of lying, election outcomes, terrorism, extremism, political conservatism/liberalism, personal health, and attitude heritability (e.g., Conway et al., 2018, Political Psychology; Conway et al, 2017, BMJ Open; Conway, Houck, Gornick, & Repke, 2017, Political Psychology;  Houck, Repke, & Conway, 2017, Journal of Policing, Intelligence, and Counter-Terrorism; Repke, Conway, & Houck, 2017, Journal of Language and Social Psychology;  Conway, Houck, Gornick, & Repke, 2016, Journal of Language and Social Psychology; Conway, Gornick, Houck, & Conway, 2015, Political Psychology; Houck, Conway, & Gornick, 2014, Political Psychology; Conway, Gornick, Burfiend, Mandella, Kuenzli, & Houck, 2012, Political Psychology; Conway et al., 2008).  

What is cognitive complexity?

Cognitive complexity is how complexly or simply people think about a particular issue. So, for example, I may think "broccoli is terrible -- I hate it." That's a pretty simple thought -- one idea about broccoli. But I may think something else about broccoli, like "broccoli has a terrible texture, but a nice flavor." (Ok, the flavor part is totally untrue, but work with me here). That's more complex -- it contains two distinct ideas about broccoli. Further, I may think “broccoli has a terrible texture and a nice flavor; but really, it’s the way the flavor and texture combine in the palate that make the unique broccoli experience.” That’s more complex still – I’m not just presenting two distinct ideas (flavor and texture), but I’m discussing how they are interrelated. Well, it turns out that almost any statement ever recorded in the history of human beings can be coded along this complexity dimension. In our lab, we most typically use the well-validated integrative complexity construct to code written or spoken statements on a 1-7 scale, as well as two new constructs of our own design (see below) known as elaborative complexity and dialectical complexity. These two new constructs help us "break down" the overall integrative complexity score into its component parts, in particular whether it emerged because someone is defending a single view complexly ("elaborative" complexity) versus actually considering multiple views in a complex way ("dialectical" complexity).

What is cognitive complexity, Part II?

Our lab has pioneered the development of the cognitive complexity construct with a new, innovative way of determining what kind of complexity we are talking about. In particular, consider the following two statements: (1) "Broccoli has a terrible texture but a great flavor." (2) "Broccoli has a terrible texture and, completely independent of that, it also has a terrible flavor." Both statements would score a 3 on the intergrative complexity scale, because both clearly discuss different dimensions associated with broccoli. Yet few would argue that the two statements are at least somewhat psychologically different: The first statement acknowledges pros and cons to broccoli, while the second offers a unilaterally negative view. In response, we have developed a system that allows to tell how much of a particular integrative complexity score is due to the first kind of thing (which we call dialectical complexity) versus the second kind of thing (which we call elaborative complexity). We have published multiple papers both validating the new system and using it to study national security phenomena such as lying, extremism, ideology, and terrorism (Repke et al., 2018, Journal of Language and Social Psychology; Houck et al., 2017, Journal of Policing, Intelligence, & Counter Terrorism; Houck et al., 2014; Political Psychology; Conway, Gornick, Burfiend, Mandella, Kuenzli, & Houck, 2012, Political Psychology; Conway, Dodds, Towgood, McClure, & Olson, 2011, Journal of Personality; Conway, Thoemmes, Allison, Towgood, Wagner, Davey, Salcido, Stovall, Dodds, Bongard, & Conway, 2008, Journal of Personality and Social Psychology).   We have further developed a natural language processing system for scoring these two additional constructs (Houck et al., 2014, Political Psychology).

What is cognitive complexity, Part III?

As mentioned earlier, we have designed a natural language processing system to automatically score complexity via the computer. This Automated Integrative Complexity (AutoIC) system scores differentiation (distinguishing different dimensions) and integration (the recognition of the interplay of differentiated dimensions) in the same hierarchical fashion as the human-scored IC system.  In our original validity paper, AutoIC showed higher overall and prospective correlations (that is, correlations on new materials) with expert human scorers than other attempts to automate the construct (average overall r = .46; average prospective r = .41; Conway et al., 2014, Political Psychology).  Further, AutoIC replicated effects from human-scored IC in the Bush/Kerry debates, Obama/McCain debates, early Christian writings, and smoking/health domains (Conway et al., 2014).  

Since the original paper, validity evidence for AutoIC has grown.  Additional studies have compared expert human scorers to AutoIC and shown continued positive results – indeed, in each case, the human-AutoIC correlation was greater than in the original Conway et al paper (Houck et al., 2018; McCullough & Conway, 2018a; Prinsloo, 2016).  Further, AutoIC has been used in multiple studies across a wide variety of domains to produce theoretically-interpretable findings, including in the domains of terrorism (Houck et al., 2017; Putra et al., 2018), personality (Conway & Woodard, 2019), fictional versus real dialogue (McCullough & Conway, 2018a), decision-making (Prinsloo, 2016), the film industry (McCullough & Conway, 2018b), video game dialogue (McCullough, 2019), fan fiction (McCullough, 2019), religion (Houck et al., 2018), and social media (McCullough & Conway, 2019).

Academics can use the system for free -- instructions and a link to the portal can be found on the "Automated Integrative Complexity" link in the lower right hand corner.