Autonomous Economic Agents Decentralized Protocol Framework
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous economic agents (AEAs) face interoperability issues and require a trusted central party for data sharing, leading to privacy concerns and inefficient training processes, especially when data variations occur across different sources.
Innovation Solution
A decentralized system with a domain-independent protocol specification language, modular software modules, and a co-learning software module on a distributed ledger, enabling AEAs to collaborate securely without sharing metadata, using micro-agents to generate and execute protocols for service requests while leveraging external machine learning models for collective learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a trusted central party is used to aggregate information into a global model, then machine learning collaboration is enabled, but privacy risks increase and control is concentrated in a single party
Solution Approach 1:
The patent extracts the central party from the machine learning collaboration system and replaces it with a decentralized network of autonomous economic agents. Each agent maintains local control over its data and models while participating in collective learning through the OEF protocol, eliminating the need for a trusted central authority and thereby reducing privacy risks.
Solution Approach 2:
The patent segments the centralized machine learning process into distributed components where multiple autonomous economic agents independently manage their own data and models. The system divides the learning task across multiple agents that collaborate through standardized protocols, allowing each agent to maintain privacy while contributing to collective intelligence.
2Ease of manufacture
If existing protocol languages are used for multi-agent systems, then implementation is simplified, but interoperability issues arise during interactions
Solution Approach 1:
The patent implements a universal protocol language within the Open Economic Framework that enables multiple autonomous economic agents with different functionalities to interact seamlessly. The protocol provides standardized interfaces and communication mechanisms that work across diverse agent types and domains, ensuring both ease of implementation and reliable interoperability.
3Object-affected harmful factors
If federated learning is used to preserve privacy, then data sharing is reduced, but the technique requires a trusted central party and may not be practical for certain applications
Solution Approach 1:
The patent enables autonomous economic agents to independently manage their own data and machine learning models without requiring a central coordinating party. Each agent autonomously participates in the co-learning process by sharing only necessary model updates or insights through the OEF protocol, thereby maintaining privacy while eliminating the need for centralized control and increasing application flexibility.
4Object-affected harmful factors
If differential privacy technique is used, then privacy-preserving solutions are provided, but significant computational resources are required leading to slow training processes
Solution Approach 1:
The patent implements a co-learning approach where autonomous economic agents share only partial information (such as model updates, insights, or aggregated statistics) rather than complete datasets or full models. This partial information sharing achieves privacy protection goals while significantly reducing computational overhead and improving training efficiency compared to full differential privacy implementations.
Data Source
AI summary
Disclosed is system enabling application of autonomous economic agents (AEAs) across problem domains. The system comprises decentralised computing network configured to implement software framework including domain-independent protocol specification language (DIPSL), protocol generator (PG), AEAs communicably coupled with micro- agent s; and external computing arrangement comprising external computing device (ECDs) that is part of distributed ledger arrangement. Micro- agent is configured to generate invocation of PG for generating protocol(s) for protocol specification (PS). ECDs are configured to receive, from micro-AEA, invocation of PG, generate insight corresponding to action, using external machine learning model and/or co-learning software module (CSM) and transmit insight to micro- agent; micro- agent configured to receive insight from ECDs and transmit metadata thereto, upon receiving metadata from micro- agent, ECDs is configured to generate inference by applying insight and metadata to external machine learning model and/or CSM, transmit inference to micro- agent, and PG is configured to generate implementation of protocol(s) to implement PS using inference and DIPSL.