E01¶


Research	R0048 — Corporate AI Training
Run	2026-04-01
Query	Q002
Source	SRC05
Evidence	SRC05-E01
Type	Reported

Corporate approaches to automation bias training — Microsoft and IBM examples

URL: https://www.lumenova.ai/blog/overreliance-on-ai-adressing-automation-bias-today/

Extract¶

Corporate training approaches to automation bias:

Microsoft: Provides "training sessions for users demonstrating scenarios where AI excels and where it might falter"
IBM Watson: Includes "detailed explanations for treatment recommendations, including data sources and logic"
Research finding: "27.7% of students demonstrated degraded decision-making skills due to their reliance on AI"
Contributing factors to overtrust: lack of AI literacy, domain expertise paradox, cognitive offloading under time pressure
Nine cognitive biases identified as relevant: automation bias, confirmation bias, ordering effects, and others
Recommended mitigations: real-time feedback, cognitive forcing functions, transparency requirements

Critically: the article addresses automation bias and overreliance mechanisms but does not connect these to sycophancy or acknowledge that AI systems may actively reinforce user expectations.

Relevance to Hypotheses¶

Hypothesis	Relationship	Strength
H1	Contradicts	Examples show adjacent-concept training exists but not sycophancy-specific
H2	Supports	Microsoft and IBM examples demonstrate automation bias awareness without sycophancy coverage
H3	Contradicts	Some training does address related concepts, preventing full H3 support

Context¶

This source provides the closest evidence to corporate training addressing sycophancy-adjacent concepts. Microsoft's "where AI might falter" training is the most directly relevant corporate training example found, but it frames the issue as AI capability limitations rather than AI behavioral tendencies.