Visualizing search behavior with adaptive discriminations

Abstract

We examined different aspects of the visual search behavior of a pigeon using an open-ended, adaptive testing procedure controlled by a genetic algorithm. The animal had to accurately search for and peck a gray target element randomly located from among a variable number of surrounding darker and lighter distractor elements. Display composition was controlled by a genetic algorithm involving the multivariate configuration of different parameters or genes (number of distractors, element size, shape, spacing, target brightness, and distractor brightness). Sessions were composed of random displays, testing randomized combinations of these genes, and selected displays, representing the varied descendants of displays correctly identified by the pigeon. Testing a larger number of random displays than done previously, it was found that the bird’s solution to the search task was highly stable and did not change with extensive experience in the task. The location and shape of this attractor was visualized using multivariate behavioral surfaces in which element size and the number of distractors were the most important factors controlling search accuracy and search time. The resulting visualizations of the bird’s search behavior are discussed with reference to the potential of using adaptive, open-ended experimental techniques for investigating animal cognition and their implications for Bond and Kamil’s innovative development of virtual ecologies using an analogous methodology.

Experimental Divergences in the Visual Cognition of Birds and Mammals

“…Such evidence is more easily requested than collected. One new method we have recently started developing is to determine which features are critical or relevant by using genetic algorithms to isolate and extract the best stimulus configurations and features as identified by the selection behavior of the birds (Cook & Qadri, 2013, 2014). Regardless of the analytic tool employed, our history with studying pigeons has shown that these are efficient and “clever” problem solvers, regularly finding unanticipated solutions to our discriminative tasks.…”

Adaptive testing of the critical features in 2D-shape discrimination by pigeons and starlings.

“…This tracking randomness permits better evaluation of the birds’ current solution while avoiding local maxima. Because the pigeons started before the starlings, their initial acquisition used tracking randomness, which we employed from other settings (Cook & Qadri, 2014). However, as described in the results below, employing this method resulted in early “genetic bottlenecks”—situations where trial variation was limited around a particular subset of gene conditions.…”

“…Finally, as the starlings started some months after the pigeons, some pitfalls were avoided. For example, the pigeons were initially trained using a modified GA procedure that mimicked those used in Cook and Qadri (2014) and subsequently all testing switched to the one used in Cook and Qadri (2013). Despite these differences, both species had an extended and comparable period of testing following their learning of the discrimination.…”

“…GAs have had increasingly widespread impact across a number of areas of science and engineering, including biology, economics, bioinformatics, robotics, and machine learning. While their application and impact in comparative psychology has been minimal, their utility for the experimental control and measurement of behavior is considerable (Bond & Kamil, 2002; Cook & Qadri, 2013, 2014). Here, the adaptive procedure allows us to sample and investigate a broader and more varied set of shapes and organizations than previously permitted using traditional fixed-stimuli displays.…”