Large language models for causal hypothesis generation in science
Towards the goal of understanding the causal structure underlying complex systems—such as the Earth, the climate, or the brain—integrating Large language models (LLMs) with data-driven and domain-expertise-driven approaches has the potential to become a game-changer, especially in data and expertise...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2025-01-01
|
| Series: | Machine Learning: Science and Technology |
| Subjects: | |
| Online Access: | https://doi.org/10.1088/2632-2153/ada47f |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832576049971462144 |
|---|---|
| author | Kai-Hendrik Cohrs Emiliano Diaz Vasileios Sitokonstantinou Gherardo Varando Gustau Camps-Valls |
| author_facet | Kai-Hendrik Cohrs Emiliano Diaz Vasileios Sitokonstantinou Gherardo Varando Gustau Camps-Valls |
| author_sort | Kai-Hendrik Cohrs |
| collection | DOAJ |
| description | Towards the goal of understanding the causal structure underlying complex systems—such as the Earth, the climate, or the brain—integrating Large language models (LLMs) with data-driven and domain-expertise-driven approaches has the potential to become a game-changer, especially in data and expertise-limited scenarios. Debates persist around LLMs’ causal reasoning capacities. However, rather than engaging in philosophical debates, we propose integrating LLMs into a scientific framework for causal hypothesis generation alongside expert knowledge and data. Our goals include formalizing LLMs as probabilistic imperfect experts, developing adaptive methods for causal hypothesis generation, and establishing universal benchmarks for comprehensive comparisons. Specifically, we introduce a spectrum of integration methods for experts, LLMs, and data-driven approaches. We review existing approaches for causal hypothesis generation and classify them within this spectrum. As an example, our hybrid (LLM + data) causal discovery algorithm illustrates ways for deeper integration. Characterizing imperfect experts along dimensions such as (1) reliability, (2) consistency, (3) uncertainty, and (4) content vs. reasoning are emphasized for developing adaptable methods. Lastly, we stress the importance of model-agnostic benchmarks. |
| format | Article |
| id | doaj-art-6605f3b5eace4ea9802d52c10afa0e72 |
| institution | Kabale University |
| issn | 2632-2153 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IOP Publishing |
| record_format | Article |
| series | Machine Learning: Science and Technology |
| spelling | doaj-art-6605f3b5eace4ea9802d52c10afa0e722025-01-31T13:28:55ZengIOP PublishingMachine Learning: Science and Technology2632-21532025-01-016101300110.1088/2632-2153/ada47fLarge language models for causal hypothesis generation in scienceKai-Hendrik Cohrs0https://orcid.org/0000-0002-2286-7487Emiliano Diaz1https://orcid.org/0000-0001-8410-6635Vasileios Sitokonstantinou2https://orcid.org/0000-0001-5506-2872Gherardo Varando3https://orcid.org/0000-0002-6708-1103Gustau Camps-Valls4https://orcid.org/0000-0003-1683-2138Image Processing Laboratory (IPL), Universitat de València , València, SpainImage Processing Laboratory (IPL), Universitat de València , València, SpainImage Processing Laboratory (IPL), Universitat de València , València, SpainImage Processing Laboratory (IPL), Universitat de València , València, SpainImage Processing Laboratory (IPL), Universitat de València , València, SpainTowards the goal of understanding the causal structure underlying complex systems—such as the Earth, the climate, or the brain—integrating Large language models (LLMs) with data-driven and domain-expertise-driven approaches has the potential to become a game-changer, especially in data and expertise-limited scenarios. Debates persist around LLMs’ causal reasoning capacities. However, rather than engaging in philosophical debates, we propose integrating LLMs into a scientific framework for causal hypothesis generation alongside expert knowledge and data. Our goals include formalizing LLMs as probabilistic imperfect experts, developing adaptive methods for causal hypothesis generation, and establishing universal benchmarks for comprehensive comparisons. Specifically, we introduce a spectrum of integration methods for experts, LLMs, and data-driven approaches. We review existing approaches for causal hypothesis generation and classify them within this spectrum. As an example, our hybrid (LLM + data) causal discovery algorithm illustrates ways for deeper integration. Characterizing imperfect experts along dimensions such as (1) reliability, (2) consistency, (3) uncertainty, and (4) content vs. reasoning are emphasized for developing adaptable methods. Lastly, we stress the importance of model-agnostic benchmarks.https://doi.org/10.1088/2632-2153/ada47fcausalitylarge language modelshypothesis generationsciencecausal discovery |
| spellingShingle | Kai-Hendrik Cohrs Emiliano Diaz Vasileios Sitokonstantinou Gherardo Varando Gustau Camps-Valls Large language models for causal hypothesis generation in science Machine Learning: Science and Technology causality large language models hypothesis generation science causal discovery |
| title | Large language models for causal hypothesis generation in science |
| title_full | Large language models for causal hypothesis generation in science |
| title_fullStr | Large language models for causal hypothesis generation in science |
| title_full_unstemmed | Large language models for causal hypothesis generation in science |
| title_short | Large language models for causal hypothesis generation in science |
| title_sort | large language models for causal hypothesis generation in science |
| topic | causality large language models hypothesis generation science causal discovery |
| url | https://doi.org/10.1088/2632-2153/ada47f |
| work_keys_str_mv | AT kaihendrikcohrs largelanguagemodelsforcausalhypothesisgenerationinscience AT emilianodiaz largelanguagemodelsforcausalhypothesisgenerationinscience AT vasileiossitokonstantinou largelanguagemodelsforcausalhypothesisgenerationinscience AT gherardovarando largelanguagemodelsforcausalhypothesisgenerationinscience AT gustaucampsvalls largelanguagemodelsforcausalhypothesisgenerationinscience |