AI Character Model Goal-Oriented Response Modification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing virtual character models lack the ability to adapt their interactions with users based on specific goals, limiting their effectiveness in achieving desired outcomes during conversations.

Innovation Solution

A computing platform that assigns goals to AI character models, modifies responses from language models to align with these goals, and prompts users to achieve those goals, enabling goal-oriented behavior in virtual character interactions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If conventional virtual character models follow specific rules and logic, then they maintain stable behavior, but they lack the ability to adapt responses based on interaction goals

Engineering Contradiction:

Improveability to adapt responses based on goalsVSAvoidcomplexity of goal management system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system segments the virtual character's behavior into distinct components: a language model handles natural language generation while a separate goal management system handles objective tracking and response modification. This segmentation allows each component to specialize, with the language model focusing on fluent text generation and the goal management system focusing on adaptive behavior, thereby achieving adaptability without overwhelming complexity

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The goal management system acts as an intermediary layer between the user's input and the language model's output. It receives the raw response from the language model, evaluates it against current goals, and modifies the response if necessary to ensure goal alignment. This intermediary approach enables goal-oriented adaptability while keeping the language model itself relatively simple and unchanged

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If virtual character models use fixed parameters throughout interactions, then they maintain consistency, but they cannot achieve specific goals during conversations

Engineering Contradiction:

Improveability to change behavior to achieve goalsVSAvoidconsistency of character behavior

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system implements dynamic parameter adjustment where the virtual character's behavior parameters are not fixed but can be modified during interactions based on goal requirements. The goal management system dynamically adjusts response characteristics, tone, and content to achieve specific objectives while the underlying character definition remains consistent, enabling adaptability without sacrificing core reliability

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes parameters of the response output from the language model based on detected goals. Instead of changing the fundamental character parameters, it modifies response parameters such as politeness level, directness, information density, and tone to align with current interaction goals, thereby achieving behavioral adaptability while maintaining character consistency

Inventive Principle:

Principle #35Parameter changes

3Productivity

If virtual characters provide generic responses, then they maintain ease of implementation, but they lack effectiveness in achieving desired outcomes

Engineering Contradiction:

Improveeffectiveness in achieving interaction goalsVSAvoidcomplexity of response modification system

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The system applies partial modification to responses rather than complete regeneration. It takes the language model's existing response and selectively modifies only the portions that need adjustment to achieve goals, rather than discarding and recreating the entire response. This partial action approach achieves goal effectiveness while keeping the system relatively simple by leveraging the language model's existing work

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The goal management system implements a feedback loop where it continuously monitors interaction state, compares current behavior against desired goals, and adjusts responses accordingly. This feedback mechanism ensures effectiveness in achieving goals by constantly evaluating and refining the virtual character's behavior based on interaction outcomes and goal progress

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12033265B2Artificial intelligence character models with goal-oriented behavior

Publication Date: 2024.07.09 THEAI INC

AI summary

Systems and methods for providing an Artificial Intelligence (AI) character model with a goal-oriented behavior are provided. An example method includes assigning, to the AI character model, a goal to be achieved during an interaction of the AI character model and a user of a client-side computing device; receiving, from the client-side computing device, a message of the user; generating, based on the message, an input to a language model configured to predict a first response to the message; providing the input to the language model to obtain the first response; modifying, based on the goal, the first response to obtain a second response; and transmitting the second response to the client-side computing device. The client-side computing device presents the second response to the user. The second response is configured to prompt the user to provide a response to achieve the goal.