Exploring the use of Bayesian networks to model noticing patterns for groups of teachers and changes in noticing patterns over time

Scores on measures of teacher noticing reflect noticing skills but provide little information on what teachers notice. In this study we explore the use of Bayesian networks, which model the relationships between variables as probabilistic dependencies, as a potentially novel and complementary measure of teacher noticing. Such models can show, for groups of teachers, what information or events they notice and how what they notice influences the noticing of other information or events. We present preliminary results from 22 second grade teachers, who participated in a larger 3-year intervention study (N = 86) as members of the treatment group. Teachers responded in writing to video clips of authentic classroom instruction and associated prompts, taken as records of their noticing. We coded the mathematical and pedagogical information or events in their responses to a single video clip for each of the three years of study. The prompt, focused on decision making, asked teachers to generate one or more mathematical questions that would extend or deepen the students’ thinking, and to explain how their question(s) would accomplish this goal. We created Bayesian networks for each year and compared the network structure and strength of the probabilistic dependencies to evaluate the networks’ usefulness in identifying noticing patterns and changes in noticing patterns over time. We found, for this group of teachers, that the Bayesian networks identified interpretable and changing noticing patterns consistent with intervention goals. Implications of our findings in light of existing noticing measures are discussed.