Has the Past Stopped Predicting: Rethinking Case-Based Learning in IA

Session ID: 21
Status: pending
Format: Paper session
Accepts Submissions: Open
Conference Stream: T7. Learning from past cases to better predict future impacts

Session Description

Impact assessment has long relied on historical case comparison — using past projects to anticipate future impacts. This assumes a stable relationship between cause and effect over time, an assumption which is increasingly strained. Climate change is shifting the physical baselines against which impacts have traditionally been measured, including rainfall intensity, storm frequency, and permafrost stability. Meanwhile, emerging extraction technologies in mining and oil and gas — deep-sea mining, direct lithium extraction, carbon capture and storage — are generating impact pathways with little or no case history to draw on.
This session explores the question: under what conditions does historical case learning still work in impact assessment, and where does it actively mislead practitioners? We are particularly interested in cases where reliance on past precedent produced false confidence, and in what has been used instead — scenario-based modelling, first-principles approaches, or adaptive monitoring with pre-committed response thresholds.
We invite presentations from practitioners, regulators, and researchers in mining, oil and gas, and related sectors addressing: cases where historical precedent held up and usefully informed prediction; cases where it failed or misled, particularly due to climate-driven baseline shift or novel technology; and/or new methods developed to replace case-based prediction where it no longer holds. Presenters will be asked to characterize which of these categories their case represents.
What should replace historical case comparison when it fails, and how does impact assessment practice and guidance need to adapt in response?

Session Summary

Historical case comparison has long guided impact assessment, but climate change and new extraction technologies in mining and oil and gas are straining that assumption. This session asks when past-case learning still works, when it misleads, and what should replace precedent when it fails.

Session Chair

Janet Blackadar
University of New Brunswick
Canada