Semantic Causal-Factor Inference from Aviation Incident Narratives Using A Variational Autoencoder with Cosine-Similarity-Based Reconstruction
Authors: AZIIDA NANYONGA1, HASSAN WASSWA, UGUR TURHAN, KEITH FRANCIS JOINER, GRAHAM WILD
Organizations: SDAIA-KFUPM Joint Research Center for Artificial Intelligence, King Fahd University of Petroleum and Minerals (KFUPM), Dhahran 31261, Saudi Arabia · School of Systems and Computing, University of New South Wales, Canberra, ACT 2600, Australia · School of Science, University of New South Wales, Canberra, ACT 2612, Australia · Capability Systems Centre, University of New South Wales, Canberra, ACT 2600, Australia
Aviation accident and incident investigations generate extensive unstructured textual information containing evidence relevant to the causes and contributing factors of safety occurrences. Automatically extracting such information is challenging because causal evidence may be distributed across long and complex investigation narratives. This study proposes a semantic causal-factor inference framework combining natural language processing with a variational autoencoder (VAE) to learn the relationship between aviation investigation narratives and expert-reported probable causes. Investigation narratives and their corresponding probable causes are transformed into numerical representations, after which the encoder maps narrative representations to a probabilistic latent space. The decoder estimates representations of the corresponding probable causes and is trained using an objective that combines Kullback-Leibler divergence with cosine-similarity-based semantic reconstruction. The framework was evaluated using 20,919 finalized U.S. National Transportation Safety Board investigation reports from 2005 to 2020. On the held-out test set, the predicted and expert-reported probable-cause representations achieved a mean cosine similarity of 0.786 (SD = 0.120). The predicted representations also yielded interpretable terms associated with causal information in the reports. The results demonstrate the potential of probabilistic latent representation learning for AI-assisted extraction of causal information from aviation safety narratives while retaining expert investigation as the basis for formal causal determination.
Figures & tables
Figure 1. Proposed framework for semantic causal-factor inference from aviation investigation narratives. The investigation narrative is preprocessed and numerically represented before being encoded into a latent representation. The decoder estimates the representation of the corresponding probable cause, and semantic similarity between the predicted and expert-reported representations is incorporated into model training.
Characteristic
Value
Data source
U.S. National Transportation Safety Board (NTSB)
Investigation status
Finalized reports
Study period
2005–2020
Number of records
20,919
Model input
Analysis narrative
Model target
Probable cause
Table 1. Characteristics of the NTSB aviation investigation dataset used in this study.
Encoder
Decoder
Layer
Filters/Units
Output Shape
Activation
Kernel Size
Parameters
Layer
Filters/Units
Output Shape
Activation
Kernel Size
Parameters
Input
–
(1200,1)
–
–
0
Input
20
(20,)
–
–
0
–
–
–
–
–
0
Dense
64
(64,)
LeakyReLU
–
1,344
–
–
–
–
–
0
Reshape
–
(1,64)
–
–
0
Conv1D
256
(134,256)
LeakyReLU
36
9,472
Conv1DTranspose
64
(3,64)
LeakyReLU
4
16,448
Conv1D
128
(17,128)
LeakyReLU
25
819,328
Conv1DTranspose
64
(6,64)
LeakyReLU
4
16,448
Table 2. Final VAE encoder and decoder architecture used in the experiments, including the number of trainable parameters in each layer.
Figure 2. Inference procedure for estimating causal information from an aviation investigation narrative. The narrative sequence si is converted into a numerical representation vi and processed by the encoder to obtain the latent representation zi . The decoder produces the predicted causal representation v^i , which is subsequently mapped back to the textual representation space to obtain interpretable causal terms s^i .
Optimizer
Learning Rate
Mean Cosine Similarity
SD
Adam
0.0100
0.623
0.126
0.0010
0.663
0.150
0.0001
0.644
0.163
Nadam
0.0100
0.663
0.156
0.0010
0.661
0.140
0.0001
0.644
0.158
Table 3. Optimizer and learning-rate evaluation during model development.
Figure 3. Distribution of cosine-similarity scores between the predicted causal representations and the corresponding expert-reported probable-cause representations for the held-out test observations.
Figure 4. Cosine-similarity scores obtained across individual observations in the held-out test subset.
Figure 5. Word-cloud visualizations of terms recovered from predicted causal representations for four selected test observations. The examples contain terms associated with (a) operational, maintenance, and environmental factors; (b) engine and mechanical components; (c) flight dynamics and landing-related conditions; and (d) inspection, aircraft-component, and damage-related information. The word clouds are provided as qualitative illustrations of the information represented by the model outputs rather than as quantitative measures of causal importance.
Figure 6. Example of the original analysis narrative and corresponding probable-cause information extracted from an NTSB investigation record before text preprocessing.
Figure 7. Processed textual representations of the analysis narrative and probable cause for the example investigation record after tokenization and removal of stop words and punctuation.
Construction accident narratives contain rich causal information, but the evidence is often implicit, long-span, and distributed. We introduce ConstructCIE, a manually annotated dataset for Causal Information Extraction from OSHA construction accident reports. The dataset uses a hierarchical schema for accident types, causal factors, sub-causal factors, and supporting evidence spans. We evaluate supervised sequence taggers and instruction-tuned LLMs in an end-to-end hierarchical extraction setting. Results show that most evaluated models achieve strong accident-type prediction and recover broad causal meaning but remain limited in precise span-level extraction. JHE generally achieves stronger exact and soft matching, while IHE sometimes achieves higher keyword F1. Error distributions vary by extraction strategy, but evidence-selection and span-boundary errors remain common. These findings show that reliable Causal Information Extraction for construction accidents requires stronger domain grounding and more accurate evidence extraction.
Hung Nguyen, Jaehoon Lee, Namgyun Kim +1
Department of Computer Science, Texas A&M University · Department of Construction Science, Texas A&M University
As AVs grow in complexity and diversity, identifying the root causes of operational failures has become increasingly complex. The heterogeneity of system architectures across manufacturers, ranging from end-to-end to modular designs, together with variations in algorithms and integration strategies, limits the standardization of incident investigations and hinders systematic safety analysis. This work examines real-world AV incidents reported in the NHTSA database. We curate a dataset of 2,168 cases reported between 2021 and 2025, representing more than 80 million miles driven. To process this data, we introduce CRASH, Cognitive Reasoning Agent for Safety Hazards, an LLM-based agent that automates reasoning over crash reports by leveraging both standardized fields and unstructured narrative descriptions. CRASH operates on a unified representation of each incident to generate concise summaries, attribute a primary cause, and assess whether the AV materially contributed to the event. Our findings show that (1) CRASH attributes 64% of incidents to perception or planning failures, underscoring the importance of reasoning-based analysis for accurate fault attribution; and (2) approximately 50% of reported incidents involve rear-end collisions, highlighting a persistent and unresolved challenge in autonomous driving deployment. We further validate CRASH with five domain experts, achieving 86% accuracy in attributing AV system failures. Overall, CRASH demonstrates strong potential as a scalable and interpretable tool for automated crash analysis, providing actionable insights to support safety research and the continued development of autonomous driving systems.
Erick Silva, Rehana Yasmin, Ali Shoker
King Abdullah University of Science and Technology, Thuwal, Saudi Arabia
Operational hazard analysis of aviation system operations must consider interactions among weather, ATC actions, airspace constraints, aircraft operations, and human factors - distinct from the functional hazard assessment applied at the aircraft-system level. We present an AI-assisted approach that generates candidate hazard scenarios from NASA's Aviation Safety Reporting System (ASRS). Given a target adverse outcome, it produces a structured hypothesis as categorical factors and a narrative scenario describing an operational event sequence consistent with the structure. Each scenario includes by a plausibility score from historical co-occurrence evidence and traceability to the most similar held-out ASRS reports. We then propose a hybrid variant, conditioning narrative generation on a structured hypothesis produced via evolutionary abduction, improving correctness and reducing variability. We evaluate multiple large language models, zero-shot versus few-shot prompting, and optional fine-tuning, measuring how prompting and model choice affect the validity and realism of the generated structures and narratives.
Cristian Mascia, Roberto Pietrantuono, Daniel Rodriguez +1
University of Naples Federico II · Naples, Italy · University of Alcal´a +1