BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Asia/Seoul
X-LIC-LOCATION:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:18871231T000000
DTSTART:19881009T020000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230103T035309Z
LOCATION:Room 323\, Level 3\, West Wing
DTSTART;TZID=Asia/Seoul:20221207T142400
DTEND;TZID=Asia/Seoul:20221207T144800
UID:siggraphasia_SIGGRAPH Asia 2022_sess256_gp_141@linklings.com
SUMMARY:Towards Relatable Explainable AI with the Perceptual Process
DESCRIPTION:Talks\n\nTowards Relatable Explainable AI with the Perceptual 
 Process\n\nZhang\n\nMachine learning models need to provide contrastive ex
 planations, since people often seek to understand why a puzzling predictio
 n occurred instead of some expected outcome. Current contrastive explanati
 ons are rudimentary comparisons between examples or raw features, which re
 main difficult to interpret, since they lack semantic meaning. We argue th
 at explanations must be more re- latable to other concepts, hypotheticals,
  and associations. Inspired by the perceptual process from cognitive psych
 ology, we propose the XAI Perceptual Processing Framework and RexNet model
  for relatable explainable AI with Contrastive Saliency, Counterfactual Sy
 nthetic, and Contrastive Cues explanations. We investigated the applicatio
 n of vocal emotion recognition, and implemented a modu- lar multi-task dee
 p neural network to predict and explain emotions from speech. From think-a
 loud and controlled studies, we found that counterfactual explanations wer
 e useful and further enhanced with semantic cues, but not saliency explana
 tions. This work pro- vides insights into providing and evaluating relatab
 le contrastive explainable AI for perception applications.\n\nRegistration
  Category: FULL ACCESS, EXPERIENCE PLUS ACCESS, TRADE EXHIBITOR\n\nLanguag
 e: ENGLISH\n\nFormat: IN-PERSON
URL:https://sa2022.siggraph.org/en/full-program/?id=gp_141&sess=sess256
END:VEVENT
END:VCALENDAR