Interpretability in Deep Time Series Models Demands Semantic Alignment

TL;DR

This paper argues that interpretability in deep time series models should focus on semantic alignment with human reasoning, emphasizing meaningful variables and temporal consistency to enhance trust and usability.

Contribution

It formalizes the concept of semantic alignment in deep time series models and proposes a framework to ensure interpretability aligns with human understanding over time.

Findings

Semantic alignment improves model trustworthiness.

Temporal consistency is crucial for interpretability.

Blueprint for designing semantically aligned models.

Abstract

Deep time series models continue to improve predictive performance, yet their deployment remains limited by their black-box nature. In response, existing interpretability approaches in the field keep focusing on explaining the internal model computations, without addressing whether they align or not with how a human would reason about the studied phenomenon. Instead, we state interpretability in deep time series models should pursue semantic alignment: predictions should be expressed in terms of variables that are meaningful to the end user, mediated by spatial and temporal mechanisms that admit user-dependent constraints. In this paper, we formalize this requirement and require that, once established, semantic alignment must be preserved under temporal evolution: a constraint with no analog in static settings. Provided with this definition, we outline a blueprint for semantically aligned deep time series models, identify properties that support trust, and discuss implications for model design.