Applying Abduction Based on LLM to Analysis and Forecasting of International Affairs for Decision Making in Business

Abstract

Since Peirce pointed out the three types of inference (deduction, induction and abduction), prior researches were conducted to apply "abduction" to algorithm of machine learning so as to develop general artificial intelligence. Nevertheless, less effort has been done yet in this regard to utilize large language model (LLM) to this direction. Furthermore, no single use case of artificial intelligence in business, which is capable of abductive reasoning, has been scientifically reported so far. This concept paper describes a research plan to apply abduction based on LLM combined with retrieval augmented generation (RAG) to analysis and forecasting of international affairs, which could give tremendous positive impacts to decision making in business. While ordinary abductive reasoning only finds out hypothesis to observed fact, the herewith proposed system could help users to make relevant future scenarios in line with the abductively selected hypothesis. To realize the concept, the key is how to pick up controversial fact of importance in vast sphere of open source, so as to kick off the entire process of abduction with the appropriate algorithm and prompts.