English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The increasing number of regulations and expectations of predictive machinelearning models, such as so called right to explanation, has led to a largenumber of methods promising greater interpretability. High demand has led to awidespread adoption of XAI techniques like Shapley values, Partial Dependenceprofiles or permutational variable importance. However, we still do not knowenough about their properties and how they manifest in the context in whichexplanations are created by analysts, reviewed by auditors, and interpreted byvarious stakeholders. This paper highlights a blind spot which, althoughcritical, is often overlooked when monitoring and auditing machine learningmodels: the effect of the reference data on the explanation calculation. Wediscuss that many model explanations depend directly or indirectly on thechoice of the referenced data distribution. We showcase examples where smallchanges in the distribution lead to drastic changes in the explanations, suchas a change in trend or, alarmingly, a conclusion. Consequently, we postulatethat obtaining robust and useful explanations always requires supporting themwith a broader context.

Do not explain without context: addressing the blind spot of model explanations