Poster #162 - Infants Prefer to Learn From Own- Over Other-Race Adults: Evidence From Computational Modeling

• In Event: 2-163 - Poster Session 08
  In Poster Session: PS 08 Section - Perceptual, Sensory, Motor

Integrative Statement

Recent studies suggest that early experiences not only determine early perceptual development, but also influence early social interactions (Quinn et al., 2018). For example, 7-month-olds followed the gaze of own-race adults, but not that of other-race adults after learning their gazes were not predictive of an event 100% of the time (Xiao et al., 2018). The findings suggest that the learning capabilities of infants during the course of experiments might be biased by their own-race face experiences. In particular, infants are more likely to learn from own-race individuals and thus follow their gaze, allowing them to overcome stochastic gaze behavior. However, this explanation is challenged by an argument that the finding might be irrelevant to learning, but may reflect a pre-existing bias to follow the gaze of own- over other-race individuals that is most evident when gaze behavior is more stochastic. Here, we used a computational modeling approach (Rescorla & Wagner, 1972) to probe whether race-related differences are present in infants’ initial predictions about the gaze behavior of own- versus other-race faces or there are differences in how infants learn to follow an adult’s gaze.
Infants (7.76 months) viewed videos (16) that depicted an adult actress (own-race/other-race) turning her head and looking at one of 4 screen corners. After the head turn, an animal appeared in one of the corners. On reliable trials, animals appeared in the same location as the looked-at location. On unreliable trials, the animal appeared in a different location. By manipulating the proportions of reliable trials, we created 3 reliability levels: 25% (non-reliable), 50% (uncertain), and 100% (perfectly reliable). With a between-subjects design, 303 infants were randomly assigned to 6 conditions (2 Races × 3 Reliabilities). Infant eye movements were recorded with an eye-tracker.
The Rescorla-Wagner model assumes that individuals adjust behavior based on the difference between predicted and actual situation (Figure 1). A behavior is maintained if the actual situation matches with the prediction and changes if it violates the prediction. We fit infant gaze-following looking time on each trial by varying the learning rate (α) and initial prediction value (P0) to minimize the Mean Square Error between P and actual gaze-following looking time.
The results showed that the reliability and race affected how infants learned from the adults (Figure 2) but we found no differences in their initial predictions based on race (P0). While we found no learning when adults were 25% reliable and robust learning when adults were 100% reliable, when the adults were 50% reliable, infants showed a significantly higher learning rate in the own-race condition than in the other-race condition (p < .05). However, we found no differences in infants’ initial predictions (P0) related to own- versus other-race faces suggesting that this bias in gaze following arises from differences in learning and not in initial looking behavior. Together with Xiao et al (2018), these findings converge to suggest that a race-based learning bias exists in infants and that asymmetrical experiences with face-race in early life may exert broad impact.

Authors

• Naiqi G Xiao, Princeton University
  Presenting Author

• Sagi Jaffe-Dax, Princeton University
  Non-Presenting Author

• Shaoying Liu, Zhejiang Sci-Tech University
  Non-Presenting Author

• Paul C Quinn, University of Delaware
  Non-Presenting Author

• Kang Lee, University of Toronto
  Non-Presenting Author

• Lauren Emberson, Princeton University
  Non-Presenting Author