Cause and Effect: Can Large Language Models Truly Understand Causality?
Open accessAbstract
With the rise of Large Language Models (LLMs), it has become crucial to understand their capabilities and limitations in deciphering and explaining the complex web of causal relationships that language entails. Current methods use either explicit or implicit causal reasoning, yet there is a strong need for a unified approach combining both to tackle a wide array of causal relationships more effectively. This research proposes a novel architecture called Context-Aware Reasoning Enhancement with Counterfactual Analysis (CARE-CA) to enhance causal reasoning and explainability. The proposed framework incorporates an explicit causal detection module with ConceptNet and counterfactual statements, as well as implicit causal detection through LLMs. Our framework goes one step further with a layer of counterfactual explanations to accentuate LLMs' understanding of causality. The knowledge from ConceptNet enhances the performance of multiple causal reasoning tasks such as causal discovery, causal identification, and counterfactual reasoning. The counterfactual sentences add explicit knowledge of `not caused by' scenarios. By combining these powerful modules, our model aims to provide a deeper understanding of causal relationships, enabling enhanced interpretability. Evaluation of benchmark datasets shows improved performance across all metrics, such as accuracy, precision, recall, and F1 scores. We also present CausalNet, a novel dataset specifically curated to benchmark and enhance the causal reasoning capabilities of LLMs. This dataset is accompanied by code designed to facilitate further research in this domain.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
These models often fail to distinguish between mere association and actual causation.
“As highlighted by recent research from DeepMind and conceptual analyses by TrueTheta [182-184], these models often fail to distinguish between mere association and actual causation.”
From AI Agents vs. Agentic AI: A Conceptual taxonomy, applications and challenges (2026) · cites this paper as [184]
The Approach section says the framework uses counterfactuals to ensure identified causal links are “not merely correlational,” and the Remark section identifies limitations in LLMs’ causal reasoning. These passages support the concern behind the claim, but they do not establish that models often mistake association for causation. (LLM, conf=0.82)
Experimental — accuracy not yet measuredLLMs, which form the cognitive core of most AI Agents, are highly effective at detecting statistical correlations within training data, but they do not truly understand cause-and-effect relationships.
“While LLMs, which form the cognitive core of most AI Agents are highly effective at detecting statistical correlations within training data, they do not truly understand cause-andeffect relationships.”
From AI Agents vs. Agentic AI: A Conceptual taxonomy, applications and challenges (2026) · cites this paper as [184]
The Introduction (page unavailable) says most LLMs lack genuine comprehension of causal mechanisms, supporting the concern about cause and effect. But the Remark section (page unavailable) also credits LLMs with strong foundational causal-reasoning abilities, so the claim that they do not truly understand causality is more absolute than the paper's account. The passages do not establish that LLMs are highly effective at detecting statistical correlations in training data or that they form the cognitive core of most AI agents. (LLM, conf=0.85)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
No references could be resolved yet.
Research works printed but not linked: 23
- Conneau, A.; Khandelwal, K.; Goyal, N.; Chaudhary, V.; Wenzek, G.; Guzmán, F.; Grave, E.; Ott, M.; Zettle- moyer, L.; and Stoyanov, V. 2019. Unsupervised cross- lingual representation learning at scale. arXiv preprint arXiv:1911.02116. · arXiv
- Devlin, J.; Chang, M.-W.; Lee, K.; and Toutanova, K. 2018. Bert: Pre-training of deep bidirectional transformers for lan- guage understanding. arXiv preprint arXiv:1810.04805. · arXiv
- Du, L.; Ding, X.; Xiong, K.; Liu, T.; and Qin, B. 2022. e-CARE: a New Dataset for Exploring Explainable Causal Reasoning. Submitted on 12 May 2022, arXiv:2205.05849. · arXiv
- He, P.; Liu, X.; Gao, J.; and Chen, W. 2020. Deberta: Decoding-enhanced bert with disentangled attention. arXiv preprint arXiv:2006.03654. · arXiv
- Jiang, A. Q.; Sablayrolles, A.; Mensch, A.; Bamford, C.; Chaplot, D. S.; Casas, D. d. l.; Bressand, F.; Lengyel, G.; Lample, G.; Saulnier, L.; et al. 2023. Mistral 7B. arXiv preprint arXiv:2310.06825. · arXiv
- Jin, Z.; Chen, Y.; Leeb, F.; Gresele, L.; Kamal, O.; Lyu, Z.; Blin, K.; Gonzalez Adauto, F.; Kleiman-Weiner, M.; Sachan, M.; and Schölkopf, B. 2023a. CLadder: Assessing Causal Reasoning in Language Models. NeurIPS 2023; updated with CLadder dataset v1.5, arXiv:2312.04350. · arXiv
- Jin, Z.; Liu, J.; Lyu, Z.; Poff, S.; Sachan, M.; Mihalcea, R.; Diab, M.; and Schölkopf, B. 2023b. Can large language models infer causation from correlation? arXiv preprint arXiv:2306.05836. · arXiv
- Jin, Z.; von Kügelgen, J.; Ni, J.; Vaidhya, T.; Kaushal, A.; Sachan, M.; and Schoelkopf, B. 2021. Causal direction of data collection matters: Implications of causal and anticausal learning for NLP. arXiv preprint arXiv:2110.03618. · arXiv
- Kıc ¸ıman, E.; Ness, R.; Sharma, A.; and Tan, C. 2023. Causal Reasoning and Large Language Models: Opening a New Frontier for Causality. arXiv preprint arXiv:2305.00050. · arXiv
- Lan, Z.; Chen, M.; Goodman, S.; Gimpel, K.; Sharma, P.; and Soricut, R. 2019. Albert: A lite bert for self- supervised learning of language representations. arXiv preprint arXiv:1909.11942. · arXiv
- Liu, Y.; Ott, M.; Goyal, N.; Du, J.; Joshi, M.; Chen, D.; Levy, O.; Lewis, M.; Zettlemoyer, L.; and Stoyanov, V. 2019. Roberta: A robustly optimized bert pretraining ap- proach. arXiv preprint arXiv:1907.11692. · arXiv
- Peña, A.; Morales, A.; Fierrez, J.; Serna, I.; Ortega-Garcia, J.; Puente, I.; Cordova, J.; and Cordova, G. 2023. Leveraging Large Language Models for Topic Classification in the Do- main of Public Affairs. Accepted in ICDAR 2023 Workshop on Automatic Domain-Adapted and Personalized Document Analysis, arXiv:2306.02864. · arXiv
- Qin, L.; Bosselut, A.; Holtzman, A.; Bhagavatula, C.; Clark, E.; and Choi, Y. 2019. Counterfactual story reasoning and generation. arXiv preprint arXiv:1909.04076. · arXiv
- Raffel, C.; Shazeer, N.; Roberts, A.; Lee, K.; Narang, S.; Matena, M.; Zhou, Y.; Li, W.; and Liu, P. J. 2020. Explor- ing the limits of transfer learning with a unified text-to-text transformer. The Journal of Machine Learning Research, 21(1): 5485-5551.
- Romanou, A.; Montariol, S.; Paul, D.; Laugier, L.; Aberer, K.; and Bosselut, A. 2023. Crab: Assessing the strength of causal relationships between real-world events. arXiv preprint arXiv:2311.04284. · arXiv
- Singh, S.; Wen, N.; Hou, Y.; Alipoormolabashi, P.; Wu, T.- L.; Ma, X.; and Peng, N. 2021. COM2SENSE: A Com- monsense Reasoning Benchmark with Complementary Sen- tences. In Findings of the Association for Computational Linguistics: ACL 2021. In Proceedings of Findings of the Association for Computational Linguistics: ACL 2021 (ACL-Findings). Contains 16 pages, 14 figures, and 11 ta- bles.
- Speer, R.; Chin, J.; and Havasi, C. 2017. ConceptNet 5.5: An Open Multilingual Graph of General Knowledge. AAAI Conference on Artificial Intelligence, 4444-4451.
- Team, G.; Anil, R.; Borgeaud, S.; Wu, Y.; Alayrac, J.-B.; Yu, J.; Soricut, R.; Schalkwyk, J.; Dai, A. M.; Hauth, A.; et al. 2023. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805. · arXiv
- Touvron, H.; Martin, L.; Stone, K.; Albert, P.; Almahairi, A.; Babaei, Y.; Bashlykov, N.; Batra, S.; Bhargava, P.; Bhosale, S.; et al. 2023. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288. · arXiv
- Weng, Y.; Zhu, M.; Xia, F.; Li, B.; He, S.; Liu, S.; Sun, B.; Liu, K.; and Zhao, J. 2023. Large language models are better reasoners with self-verification. In Findings of the Associ- ation for Computational Linguistics: EMNLP 2023, 2550- 2575.
- Zečević, M.; Willig, M.; Dhami, D. S.; and Kerst- ing, K. 2023. Causal parrots: Large language models may talk causality but are not causal. arXiv preprint arXiv:2308.13067. · arXiv
- Zhang, C.; Bauer, S.; Bennett, P.; Gao, J.; Gong, W.; Hilmkil, A.; Jennings, J.; Ma, C.; Minka, T.; Pawlowski, N.; and Vaughan, J. 2023. Understanding Causality with Large Language Models: Feasibility and Opportunities. arXiv preprint arXiv:2304.05524. · arXiv
- Zhuang, Z.; Chen, Q.; Ma, L.; Li, M.; Han, Y.; Qian, Y.; Bai, H.; Feng, Z.; Zhang, W.; and Liu, T. 2023. Through the Lens of Core Competency: Survey on Evaluation of Large Language Models. arXiv preprint arXiv:2308.07902. · arXiv
Cited by
Sorted by citations
AI Agents vs. Agentic AI: A Conceptual taxonomy, applications and challenges
Sapkota, R., Roumeliotis, K. I., & Karkee, M. (2026) · Information Fusion · doi:10.1016/j.inffus.2025.103599 · Cited by 244 · see citing works
AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challenges
Sapkota, R., Roumeliotis, K. I., & Karkee, M. (2025) · SuperIntelligence - Robotics - Safety & Alignment · doi:10.70777/si.v2i3.15161 · Cited by 186 · see citing works
- Authors
- Ashwani, S., Hegde, K., Reddy Mannuru, N., Singh Sengar, D., Jindal, M., Chaitanya Rao Kathala, K., Banga, D., Jain, V., & Chadha, A.
- Affiliations
- Carnegie Mellon University (US)
- Rensselaer Polytechnic Institute (US)
- Stanford University (US)
- Venue
- Proceedings of the AAAI Symposium Series
- Published
- 2024-11-08
- Volume
- 4
- Issue
- 1
- Pages
- 2-9
- ISSN
- 2994-4317 (ISSN-L)
- OpenAlex
- W4404188625
- Cited by
- 18
- References
- 0
- Citations by year
- 2026: 7 · 2025: 10 · 2024: 1
- FWCI
- 5.20
- Type
- conference paper
- Language
- en
- Open access
- diamond
- doi.org
- Locations
- Proceedings of the AAAI Symposium Series (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Logic, Reasoning, and Knowledge
- Keywords
- large language models, causal reasoning, counterfactual reasoning, ConceptNet, counterfactual explanations, causal discovery, causal identification, LLM interpretability
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
