AI2030 Responsible AI Tools Marketplace

Key Takeaways

Data quality issues can be detected through explanation inconsistencies: The paper demonstrates that when explanations from predictive models contradict each other across different settings, this often signals underlying data problems rather than model failures—providing a practical diagnostic tool for identifying data issues before they compromise business decisions.
Technical choices significantly shape AI explanations: The research reveals that different preprocessing decisions and model configurations in predictive process monitoring produce substantially different explanations for the same predictions, meaning organizations cannot assume explanations are objective—they reflect the specific technical pathway chosen.
Explainability should be integrated from the start, not added later: By treating explainability as a fundamental component throughout the predictive process monitoring pipeline rather than an afterthought, organizations can catch data and methodological problems earlier and build more trustworthy systems.
Multiple explanation methods should be compared: The framework constructed in this research emphasizes that relying on a single explanation technique may mask important insights; comparing explanations across different XAI methods helps reveal whether inconsistencies stem from the data, the model, or the explanation method itself.
Trust in predictions requires understanding the full pipeline: Users cannot meaningfully trust predictive systems by examining explanations alone—they must understand how data was prepared, which features were selected, and how the model was configured, as these upstream choices fundamentally shape both predictions and their explanations.

Abstract

Predictive business process monitoring (PPM) has been around for several years as a use case of process mining. PPM enables foreseeing the future of a business process through predicting relevant information about how a running process instance might end, related performance indicators, and other predictable aspects. A big share of PPM approaches adopts a Machine Learning (ML) technique to address a prediction task, especially non-process-aware PPM approaches. Consequently, PPM inherits the challenges faced by ML approaches. One of these challenges concerns the need to gain user trust in the predictions generated. The field of explainable artificial intelligence (XAI) addresses this issue. However, the choices made, and the techniques employed in a PPM task, in addition to ML model characteristics, influence resulting explanations. A comparison of the influence of different settings on the generated explanations is missing. To address this gap, we investigate the effect of different PPM settings on resulting data fed into an ML model and consequently to a XAI method. We study how differences in resulting explanations may indicate several issues in underlying data. We construct a framework for our experiments including different settings at each stage of PPM with XAI integrated as a fundamental part. Our experiments reveal several inconsistencies, as well as agreements, between data characteristics (and hence expectations about these data), important data used by the ML model as a result of querying it, and explanations of predictions of the investigated ML model.

32 pages, 10 figures, 1 table

Key Takeaways

Data quality issues can be detected through explanation inconsistencies: The paper demonstrates that when explanations from predictive models contradict each other across different settings, this often signals underlying data problems rather than model failures—providing a practical diagnostic tool for identifying data issues before they compromise business decisions.
Technical choices significantly shape AI explanations: The research reveals that different preprocessing decisions and model configurations in predictive process monitoring produce substantially different explanations for the same predictions, meaning organizations cannot assume explanations are objective—they reflect the specific technical pathway chosen.
Explainability should be integrated from the start, not added later: By treating explainability as a fundamental component throughout the predictive process monitoring pipeline rather than an afterthought, organizations can catch data and methodological problems earlier and build more trustworthy systems.
Multiple explanation methods should be compared: The framework constructed in this research emphasizes that relying on a single explanation technique may mask important insights; comparing explanations across different XAI methods helps reveal whether inconsistencies stem from the data, the model, or the explanation method itself.
Trust in predictions requires understanding the full pipeline: Users cannot meaningfully trust predictive systems by examining explanations alone—they must understand how data was prepared, which features were selected, and how the model was configured, as these upstream choices fundamentally shape both predictions and their explanations.

Abstract

32 pages, 10 figures, 1 table

Explainability of Predictive Process Monitoring Results: Can You See My Data Issues?

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links

Explainability of Predictive Process Monitoring Results: Can You See My Data Issues?

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links