- Number
- 20250390700
- Published
- 2025-12-25
- Filed
- 2025-08-26
- Assignee
- Disney Enterprises, Inc.
- Inventors
- Scanlon; Raymond J. et al.
- CPC
- G06F40/35; G06N3/006; G06T13/40; G06T19/006; G06T19/20
- Verdict
- Low Notable software
- Source
- Google Patents · FreePatentsOnline
The keeper's note
Context-based interactive social-agent (character) interaction technique (continuation).
Abstract
A system for performing context-based management of social agent interactions includes processing hardware and a memory storing a software code. The processing hardware executes the software code to detect the presence of an interaction partner, identify a present state of an interaction with the interaction partner, and to determine, based on the present state, a first score for each of multiple interactive expressions for use in initiating or continuing the interaction. The processing hardware further executes the software code to predict a state change of the interaction based on each of the interactive expressions to provide multiple predicted state changes corresponding respectively to the multiple interactive expressions, to determine, using the predicted state changes, a second score for each of the interactive expressions, and to select, using the first scores and the second scores, at least one of the interactive expressions to initiate or continue the interaction.
Background
BACKGROUND
Advances in artificial intelligence have led to the development of a variety of devices providing dialogue-based interfaces that simulate social agents. However, conventional dialogue interfaces typically project a single synthesized persona that tends to lack character and naturalness. In addition, the dialog interfaces provided by the conventional art are typically transactional, and indicate to a user that they are listening for a communication from the user by responding to an affirmative request by the user.
In contrast to conventional transactional social agent interactions, natural communications between human beings are more nuanced, varied, and dynamic. That is to say, typical shortcomings of conventional social agents include their inability to engage in natural, fluid interactions, their inability to process more than one statement or question concurrently, and their inability to repair a flaw in an interaction, such as a miscommunication or other conversation breakdown. Moreover, although existing social agents offer some degree of user personalization, for example tailoring responses to an individual user's characteristics or preferences, that personalization remains limited by their fundamentally transactional design, which makes it unnecessary for conventional social agents to remember more than a limited set of predefined keywords, such as user names and basic user preferences.
Claims
21. A system for providing context-based social agent interaction, the system comprising: a social agent, the social agent (i) including one or more mechanical actuators, or (ii) being instantiated in a speaker or on a display; one or more sensors; a processing hardware; and a memory storing an interaction manager software code; wherein the processing hardware is configured to execute the interaction manager software code to: receive input data from the one or more sensors; identify, based on the received input data, a present state of an interaction with an interaction partner, wherein the present state includes a flaw in the interaction; determine, based on the present state including the flaw, a first score for a first interactive expression and a first score for a second interactive expression; predict, based on the first interactive expression, a first state change of the interaction; determine, using the first predicted state change, a second score for the first interactive expression; predict, based on the second interactive expression, a second state change of the interaction; determine, using the second predicted state change, a second score for the second interactive expression; select, using the first score and the second score for the first interactive expression and the first score and the second score for the second interactive expression, one of the first interactive expression or the second interactive expression to repair the flaw in the interaction; and control the social agent to produce an output corresponding to the selected one of the first interactive expression or the second interactive expression. ||
31. A method for use by a system to provide context-based social agent interaction, the system including one or more sensors and a social agent (i) having one or more mechanical actuators, or (ii) being instantiated in a speaker or on a display, the method comprising: receiving input data from the one or more sensors; identifying, based on the received input data, a present state of an interaction with an interaction partner, wherein the present state includes a flaw in the interaction; determining, based on the present state including the flaw, a first score for a first interactive expression and a first score for a second interactive expression; predicting, based on the first interactive expression, a first state change of the interaction; determining, using the first predicted state change, a second score for the first interactive expression; predicting, based on the second interactive expression, a second state change of the interaction; determining, using the second predicted state change, a second score for the second interactive expression; selecting, using the first score and the second score for the first interactive expression and the first score and the second score for the second interactive expression, one of the first interactive expression or the second interactive expression to repair the flaw in the interaction; and controlling the social agent to produce an output corresponding to the selected one of the first interactive expression or the second interactive expression.