System and method for estimating confidence and implementing metacognitive abilities in artificial intelligence systems
Patent Information
- Application Number
- HK42026125639
- Authority / Receiving Office
- HK · HK
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2026-07-02
- Publication Date
- 2026-09-11
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
(12) UK Patent Application (19) GB (11) 2700827 (13) A (43) Date of A Publication 18.03.2026 (21) Application No: GB2509759.3 (22) Date of Filing: 19.06.2025 (30) Priority Data: (31) 10202401824 (32) 21.06.2024 (33) SG (71) Applicant(s): Standard Chartered Bank, Singapore Branch 8 Marina Boulevard, #27-01 Marina Bay Financial Centre, Singapore, Singapore, 018981 (72) Inventor(s): Tom Pfaff, Craig Turrell (74) Agent and / or Address for Service: Marks & Clerk LLP 40 Torphichen Street, Edinburgh, EH3 8JB (51) INT CL: G06N5 / 045 (2023.01) G06Q40 / 02 (2023.01) G06Q40 / 03 (2023.01) (56) Documents Cited: None (58) Field of Search: Other: No Search Performed 17(5)(b) (54) Title of the Invention: System and method for estimating confidence and implementing metacognitive abilities in artificial intelligence systems Abstract Title: System and method for estimating confidence and implementing metacognitive abilities in artificial intelligence system (57) A confidence score, specific to an output from an AI system, is generated. The confidence score is generated by using a language model to process an input query to generate an initial output. The processing involves computing a plurality of confidence estimates for the initial output, wherein each of the estimates is computed based on distinct computational techniques, and aggregating multiple confidence estimates based on game-theoretical approaches derived from Nash Embedding Theorem. This aggregated confidence score is used to adjust system parameters.Also disclosed, a system comprising a data acquisition module which receives feedback corresponding to an output generated by a language model, a cognitive reasoning module which evaluates the reasoning process of the language model to generate an assessment based on the feedback, a process adjustment module which adjusts the reasoning module based on the assessment, and a refinement module which iteratively refines the reasoning process based on subsequent updates of the assessment until a performance threshold is met. Also disclosed, a model configuration module which defines a structured framework of a multi-agent system wherein each component is assigned tasks related to processing outputs generated by the AI system, and an agent coordination module which manages interactions and synchronises data flows among agents.[Figure 2] Abstract This invention provides a method for generating confidence scores for the output of an artificial intelligence system. The generation process of this confidence score is as follows: First, an input query is processed using a language model to generate an initial output; second, multiple confidence estimates of the initial output are calculated, each estimate based on a different computational technique; finally, the multiple confidence estimates are aggregated using a game theory method derived from Nash embedding theorem. The aggregated confidence score is used to adjust system parameters. This invention also discloses a system comprising: a data collection module for receiving feedback corresponding to the output generated by the language model; a cognitive reasoning module for evaluating the reasoning process of the language model and generating evaluation results based on the feedback; a process adjustment module for adjusting the reasoning module based on the evaluation results; and an improvement module for iteratively updating the reasoning process based on the evaluation results until a performance threshold is reached. This invention also discloses a model configuration module that defines a structured framework for a multi-agent system, wherein each component is assigned a task related to processing the output generated by the artificial intelligence system; and an agent coordination module that manages the interaction between agents and synchronizes data streams. [Figure 2].
Claims
1. A method for generating a confidence score specific to an output from an artificial intelligence system comprising:processing an input query by a language model to obtain an initial output;computing a plurality of confidence estimates for the initial output, wherein each confidence estimate is computed based on a distinct computational technique;aggregating the confidence estimates based on a game-theoretical approach to generate an aggregated confidence score for the initial output, andapplying the aggregated confidence score to adjust parameters of an artificial intelligence system,wherein the game-theoretic approach is derived from a Nash Embedding Theorem, which optimizes the aggregation of the confidence estimates by evaluating an interdependency among the plurality of confidence estimates.
2. The method of claim 1, wherein the adjusted parameters of the artificial intelligence system are based on model performance metrics corresponding to a level of confidence determined by the aggregated confidence score.
3. The method of any one of the preceding claims, wherein the distinct computational technique is selected from a group comprising: verbalized confidence assessments, token probability analysis, prompt entropy measurements, semantic output clustering, selfconsistency diagnostics; and hidden state divergence metrics.
4. The method of any one of the preceding claims, further comprising using a structured data format to encode the confidence estimates and the aggregated confidence score, wherein the format is compatible with a standardized metadata representation.
5. The method of any one of the preceding claims, further comprising refining the game-theoretical approach based on feedback derived from system performance to enhance future confidence estimations.
6. The method of any one of the preceding claims, wherein the distinct computational technique includes calculating token probabilities by extracting SoftMax probabilities for each token from an output distribution of the language model and calculating an overall sequence probability as a product of the token probabilities.
7. The method of any one of the preceding claims, wherein the distinct computational technique includes determining a prompt entropy by prompting the Al system with multiple label options, calculating an entropy of a resulting probability distribution corresponding to the label options, and mapping entropy values to the confidence estimates based on a predetermined function.
8. The method of any one of the preceding claims, wherein the distinct computational technique includes analyzing a verbalized confidence by pattern matching to identify statements of confidence in an output of an Al system and assigning numerical scores to the identified statements based on a predefined mapping.
9. The method of any one of the preceding claims, wherein the distinct computational technique includes clustering semantic outputs by generating a plurality of output sequences, computing semantic embeddings for the output sequences, and using metrics to quantify a clustering quality.
10. A system for information processing, comprising:a data acquisition module configured to receive feedback corresponding to one or more outputs generated by a language model;a cognitive reasoning module configured to evaluate the reasoning process of the language model, emulate cognitive functions including metacognitive processes, and generate an assessment based on an analysis of the received feedback, wherein the assessment includes classifying the one or more outputs into components, assigning quality scores for each component, and identifying an improvement corresponding to the one or more outputs;a process adjustment module coupled to the cognitive reasoning module, configured to adjust the reasoning process of the language model based on the assessment; anda refinement module coupled to the process adjustment module, configured to iteratively refine the reasoning process based on subsequent updates to the generated assessment until a performance threshold is met.
11. A method for information processing, comprising:receiving feedback corresponding to one or more outputs generated by a language model;using a reasoning module to evaluate the reasoning process of the language model, emulate cognitive functions including metacognitive processes, and generate an assessment based on an analysis of the received feedback, wherein the assessment includes classifying the one or more outputs into components, assigning quality scores for each component, and identifying an improvement corresponding to the one or more outputs;adjusting the reasoning process of the language model based on the assessment; anditeratively refining the reasoning process based on subsequent updates to the generated assessment until a performance threshold is met.
12. The method of claim 11, wherein the classifying of the one or more outputs into components includes applying a modularized analysis that quantifies evaluation metrics for each component based on relevance, coherence, factual accuracy, and completeness13. The method of claim 11 or 12, wherein the assigning of quality scores for each component utilizes techniques such as token probabilities, prompt entropy, and semantic output clustering to calibrate the confidence scoring.
14. The method of any one of claims 11-13, wherein the identifying of an improvement includes employing a counterfactual reasoning process to evaluate potential alternative outcomes and their impacts on the reasoning process of the language model.
15. The method of any one of claims 11-14, wherein the reasoning module further comprises a sub-module for generating detailed logs of each iteration in the refinement process, which includes recording changes to reasoning strategies and their effects on output quality.
16. A system for coordinating operations, comprising:a model configuration module configured to define a structured framework of a multiagent system, wherein each component of the multi-agent system is assigned tasks related to processing outputs generated by an artificial intelligence (Al) system; andan agent coordination module configured to manage interactions and synchronize data flow among agents of the system based on roles or dependencies corresponding to each agent within the structured framework and utilize mechanisms for coordination and synchronization of the agents of the system.
17. The system of claim 16, wherein the model configuration module is further configured to utilize Petri nets to define the structured framework, wherein the Petri nets specify the roles or dependencies of each agent within the multi-agent system.
18. The system of claims 16 or 17, wherein the agent coordination module utilizes a conflict resolution strategy to manage data flow among the agents when multiple agents access data resources simultaneously.
19. The system of any one of claims 16-18, wherein each agent within the multi-agent system is configured to generate and send feedback regarding their task execution to the model configuration module, which uses the feedback to refine task assignments in subsequent operations.
20. The system of any one of claims 16-19, wherein the model configuration module assigns the tasks to agents based on a dynamic assessment of the operational load and performance metrics of each agent.
21. A method for coordinating operations within a multi-agent system, comprising:implementing a structured framework of a multi-agent system, wherein each component of the multi-agent system is assigned specific tasks related to processing outputs generated by an artificial intelligence (Al) system;managing interactions and synchronizing data flow among agents of the multi-agent system based on roles or dependencies corresponding to each agent within the structured framework; andutilizing mechanisms for coordination and synchronization of the agents of the system.
