It has been almost thirty years since the publication by Townsend & Nozawa (JMP; 1995) introducing systems factorial technology (SFT). Although that new-born paper has enjoyed a healthy citation rate, the most promising outcome has been the conversion of a significant number of the best mathematical and experimental psychologists to providing laboratory applications as well as theoretical and methodological extensions of the basic theory and its suites of toolboxes. Today, I apologize in advance to the presence here of a number of the core set of super-experts in SFT, who may be in danger of death-by-boredom.  Hopefully, there may be a few surprising or novel titbits even for those veterans. Also, although this conference’s centroid is the extremely important and topical subject of configural perception and cognition, to which I might add, has enjoyed many fruits issuing from SFT, my inclination is, candidly, that it is time to take stock of the original foundations of SFT, complemented by extensions and advances since 1995.  Due to time constraints, I must abstain from the many, many beautiful experimental applications of SFT over the past 3 decades in order to focus on the foundations and theoretical extensions of the (as we say) theory-driven-technology that is SFT. 

Complex tasks may require the division of labour across multiple team members. Yet assigning multiple agents to collaborate does not guarantee efficiency. Miscommunication, or constrains on resources may hamper the performance of the team, compared with what one might expect based on the individual performance of each operator alone. In this study we tested the performance of human-human and human-bot teams in an arcade-like computer game. Two players (a dyad) controlled horizontally moving paddles and had to prevent bouncing balls from hitting the virtual floor. 

We examined their performance, and behavioural patterns in three conditions: separate, where they operated individually to maximise their own personal score while ignoring the other player; collaborative, and competitive. In another set of experiments we paired human players with a bot. Behaviour of one bot-type was driven by reinforcement learning. Another bot type was loosely based on principles of ideal observer. We discuss the differences between performance, and behaviour, in the human-human and human-bot conditions.

    Melanoma is highly prevalent in the general population, with nearly 1 in 28 people being diagnosed in their lifetime and rates continuing to rise.  Fortunately, early detection has a large effect on survival rates.  Initial detection of potential melanoma often relies on perceptual expertise, whether in support of the ABCDE rule in use by the general population or in assessment by an expert dermatologist.  We suggest that configural features are critical to visual evaluation of skin lesions.  I will give an overview of our research on how novices perceive skin lesions when focused on a subset of the ABCDE diagnostic dimensions (asymmetry, boarder irregularity, and color variation).  In one study, we demonstrate that people perceive valuable information for identifying melanoma beyond that which is extracted by state-of-the-art computer vision algorithms. In the next study, we show how, even the ABC dimensions are not perceived independently.  Finally, we demonstrate the effect of perceptual expertise training, whether on ABC features individually, or on higher order features extracted from deep convolutional neural networks, influences skin lesion perception.

This talk revisits the application of Systems Factorial Technology (SFT) to understand the cognitive processes behind facial perception. Initially, I will review previous work using SFT, notably the survivor interaction contrast functions (SIC), which critically test the properties of perceptual and cognitive processes involved in facial perception. These properties include processing order (serial, parallel), search scope (terminating, exhaustive), and process dependency. Results derived from facial categorization tasks will be presented.

Building on this foundation, a novel computational model has been developed. MSPN serves both as a theoretical development tool and an exploratory tool for model validation or falsification. Historically, face processing research has been method-driven rather than theory-driven. Developments in methodologies, such as the part-to-whole and composite faces paradigms, have significantly shaped the field by providing logical foundations for understanding facial perception processes. However, recent advancements in theoretical tools like MSPN offer unprecedented detail in exploring facial perception in these paradigms.  This study leverages MSPN to establish convergent validity by examining outcomes from these two dominant paradigms. By doing so, we aim to provide additional evidence that either supports or challenges the findings revealed by MSPN. This exploration has the potential to question conventional approaches to facial perception, introducing a paradigm shift in our understanding of this complex cognitive process.