110 likes | 125 Views
Explore how an agent's gaze influences user impressions, using a gaze model to control factors like warmth, dominance, and friendliness. The study delves into gaze parameters and their impact on forming impressions in virtual interactions.
E N D
Messages Embedded in Gaze of Interface Agents: Impression Management With Agent's Gaze Atsushi Fukayama et. al. Proceedings of CHI 2002 Presentation by Kyle Johnsen
What are you talking about? • The impression of a person is our image of him that is built up in our mind from various cues, such as his utterance, behavior, appearance, and reputation.
What are you talking about? • From psychological studies • Gaze has the following functions in a conversation • monitoring – literal meaning of gaze (ie, someone is looking at something) • Regulatory – Control conversation flow • Expressive – convey some type of information • Length of eye contact used to maintain a comfortable level of intimacy [Argyle, Dean] • Continuous gaze highly rated in the impression measures of “liking/ evaluation” and “activity/ potency” [Argyle, Lefebvre, Cook]
Hypothesis • Can an embodied agent control the impression the user has of them with just an eye gaze model?
Gaze Model • Total Gaze • The percentage of time the agent looks at the user • Mean Gaze time • The mean amount of time the agent looks at the user • Gaze Points while averted • Where the agent looks when not looking directly at the user
Gaze Model • Gaze Parameters Values • Amount of Gaze : • R− =0.25, R0 =0.5, R+ =0.75, R++ =1.0 • Mean Duration of Gaze : L [ms] • L− = 500, L0 = 1000, L+ = 2000 • Gaze Points while Averted : • P⊂{(x, y)|−∞≤x, y ≤∞} • P0 = {(x, y)|−1.2 ≤x, y≤1.2}, • PH = {(x, y)|−1.0 ≤ x ≤1.0, 0.7 ≤ y ≤1.1}, • PL= {(x, y)|−1.0 ≤ x ≤1.0, −0.9 ≤ y ≤−1.3}, • PR= {(x, y)|−2.0 ≤ x ≤−1.2, −0.4 ≤ y ≤0.4
extrovert – introvert considerate – inconsiderate unassured –assured coordinative – exclusive cold– warm strong – weak selfish – responsible strict – tolerant successful – unsuccessful thoughtless – thoughtful attractive – unattractive careless– careful friendly – unfriendly unfaithful – faithful rude –respectful close – distant stubborn – flexible sociable – unsociable attentive – inattentive lazy – diligent Variables
Factor Analysis • Factor 1 (Friendliness) • friendly(0.81), warm (0.77), sociable(0.75), tolerant (0.67), flexible (0.58), attentive (0.57), coordinative (0.56) • Factor 2 (Dominance) • assured (0.84), strong (0.83), successful (0.70), responsible (0.61), careful (0.59)
Results • R values closer to R0 yield higher ratings in impression measures related to “like/dislike”. • Partially Supported • Larger L values yield higher ratings in impression measures related to “potency”. • Partially Supported
Results • L values smaller than L0 yield lower ratings in impression measures related to “strong”. • Supported • PL yields higher ratings in impression measures related to “warm”. • Not Supported • PL yields lower ratings in impression measures related to “strong”. • Supported
Conclusions • Gaze parameters can induce impressions reliably using a simple gaze model • What can we do with this? • Can we fit the user to this model so the avatar can form an impression of them? • Can head orientation be used instead of eye gaze as a rough estimate?