A unified account of gaze following

Abstract

Abstract-Gaze following, the ability to redirect one's visual attention to look at what another person is seeing, is foundational for imitation, word learning, and theory-of-mind. Previous theories have suggested that the development of gaze following in human infants is the product of a basic gaze following mechanism, plus the gradual incorporation of several distinct new mechanisms that improve the skill, such as spatial inference, and the ability to use eye direction information as well as head direction. In this paper, we offer an alternative explanation based on a single learning mechanism. From a starting state with no knowledge of the implications of another organism's gaze direction, our model learns to follow gaze by being placed in a simulated environment where an adult caregiver looks around at objects. Our infant model matches the development of gaze following in human infants as measured in key experiments that we replicate and analyze in detail.Index Terms-Adaptive systems, artificial intelligence, autonomous mental development, behavioral science, cognition, cognitive science, computational and artificial intelligence, cybernetics, emergent phenomena, intelligent systems, learning systems, multiagent systems, systems, man, and cybernetics.

When and where do infants follow gaze?

“…For example, infants respond more robustly when gaze-cues are accompanied by pointing cues; they respond first to head turns and only much later to eye-direction per se; they respond more to large head turns than small ones. All of these effects are consistent with general perceptual-motor learning processes (Jasso et al, 2012;Paulus, 2011).…”

“…Despite this limitation, the study arguably offers the strongest evidence concerning infants' ability to follow gaze cues per se, even when such a response is reinforced. The results, along with other available evidence, strongly support the view that gaze-following emerges after 6 months, and develops slowly from 9 to 36 months (Butterworth & Jarrett, 1991;Deák et al, 2000;Doherty et al, 2009), likely due to protracted learning processes working on variable, complex input (see Jasso et al, 2012;Nagai, Hosoda, Morita, & Asada, 2003;Triesch et al, 2006).…”

“…Although this contributes to the effect, the rest of the effect has not been explained. One possibility is that as infants get better at maintaining representations of an adult's head-and eye-direction when they have turned away from the adult, they become better able to reconcile the represented head/eye direction with the locations of distal objects (see Doherty et al, 2009;Jasso et al, 2012). Another contextual factor that seems to affect gazefollowing is clutter, or the concentration of visual distractors (Wass, 2014).…”

Is ostension any more than attention?

“…Thus even if infants have learned to follow gaze based on previous observations of head and eye movements to interesting sights, they still additionally require an attention-grabbing cue to do so. However, we have to highlight the fact that in contrast to research presented in9 and10, our experiments took place in laboratory conditions, in which the actors did not establish any natural interaction with children. It is thus possible that in such controlled conditions and without any history of interaction, infants need more cues to respond in a gaze following situation.…”

“…The computational model of the emergence of gaze following skills in infant-caregiver interactions910 is based on the idea that “infants learn gaze following because they discover that monitoring their caregiver's direction of gaze allows them to predict where interesting visual sights occur”9 (p. 128). Thus in this framework it is the infant's prior interactions that are crucial for the infant to form an expectation about the adult's gaze direction being linked to a target object.…”