Summary of Counterfactual Reasoning with Knowledge Graph Embeddings, by Lena Zellinger et al.
Counterfactual Reasoning with Knowledge Graph Embeddings
by Lena Zellinger, Andreas Stephan, Benjamin Roth
First submitted to arxiv on: 11 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper links knowledge graph completion and counterfactual reasoning by introducing a new task called CFKGR. It models original world states as knowledge graphs, hypothetical scenarios as added edges, and plausible changes as inferences from logical rules. The authors create benchmark datasets containing diverse scenarios with plausible changes and facts to be retained. They develop COULDD, a method for adapting existing KGEs given a hypothetical premise, and evaluate it on the benchmark. Results show that KGEs learn patterns without explicit training and adapt well to plausible counterfactual changes. However, they struggle to recognize changes not following learned rules. In contrast, ChatGPT mostly outperforms KGEs in detecting plausible changes but has poor knowledge retention. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper connects two areas: knowledge graph completion and counterfactual reasoning. It uses a new task called CFKGR. The authors model original world states as graphs, hypothetical scenarios as added edges, and plausible changes as logical rule inferences. They create benchmark datasets with diverse scenarios and facts to be retained. A method called COULDD adapts existing KGEs given a premise, and the paper evaluates it on this data. It shows that KGEs learn patterns without training and can adapt to counterfactual changes. However, they struggle with changes not following rules. ChatGPT does better at detecting changes but retains less knowledge. |
Keywords
* Artificial intelligence * Knowledge graph
