This seminar discussed tricks of the trade and methods that make patient characters that appear to be living. We will cover methods to achieve biological fidelity, interactivity, graphics and flow with the goal to introduce participants to techniques that deliver the appearance of active biology, a sense of urgency and responsiveness to game choices.
-State of the Living – Medical Games & Lifelike Patients - Thomas Talbot, USC Institute for Creative Technologies
1. VIRTUAL PATIENTS
Important aspects for
lifelike characters:
Spoken Dialogue
Thomas B. Talbot, MD, MS, FAAP
Principal Medical Expert
Institute for Creative Technologies
University of Southern California
Talbot@ict.usc.edu
Associate Research Professor of Medical Education
Keck School of Medicine of USC
Serious Play Conf July 10th, 2018
6. AGENDA
• Focus on interactive dialogue in games
• Conversational Simulations
• Dialogue agents and conversation design patterns
• Why this is important for games of the future…
30. NLRA CONVERSATIONS
• Natural Language Random Access
• Natural Language Understanding (NLU) AI Dependent
– Requires training language
– Probabilistic, not deterministic
• It is not possible to have a full scope human conversation
• Success depends on CONSTRAINTS
– Constrain domain: narrower means better performance
– Constrain initiative: exchanging initiative difficult
– Constrain scope: Go broad or deep, not both
• Lots of dialogue interaction patterns
• Possible to mix choice-based with NLRA
31. USC Standard Patient
NLRA
• 92%+ NLU Performance
• Voice Recognition decreases
NLU performance
• Strong Training Effect
• Time/Choice constraints on
users IMPROVES
performance
• People say weird things to
computers