AI and Skill Decay: How AI Is Changing Learning and Expertise

Podcast snapshot

The episode examines whether AI tools in the workplace threaten our skill and knowledge by offloading cognitive work. Host Kim Mills speaks with Dr. Brooke McNamara about cognitive offloading, retrieval practice, and how to use AI in ways that support learning rather than erode expertise. Through examples from medicine, radiology, and elite performance, the conversation also offers practical tips for trainees and professionals to stay sharp while benefiting from AI.

• Cognitive offloading may reduce skill depth if AI does the thinking for you
• Retrieval practice and self testing boost learning with AI support
• Use AI as a check, not a crutch, and test performance without AI
• Differences between novices and experts in adapting to AI

Overview

This episode from the Speaking of Psychology series features host Kim Mills in conversation with Dr. Brooke McNamara, an expert on skill acquisition and performance. The central question is how widespread use of AI in professional tasks might alter the way people learn, remember and develop expertise. The discussion situates AI within a long history of cognitive offloading, from ancient writing to GPS and calculators, and explores whether AI represents a qualitatively new challenge for learning and memory.

Historical context and cognitive offloading

The host and guest walk through the idea that people have always offloaded tasks to technology, which can change how we learn. They discuss examples such as Plato worrying about memory with writing, the shift in navigation with GPS, and the impact of calculators on math. The key concept is cognitive offloading, and the debate centers on when offloading hinders versus helps learning, depending on how people engage with AI rather than simply using it as a source of answers.

What is known about AI and skill decay

The conversation highlights that it is still early to know the full impact of AI on skill decay. Evidence from initial studies is mixed but suggests that certain professional skills may decay when not practiced directly. For example, endoscopists using AI to detect polyps showed a drop in detection performance when the AI tool was removed. Basic life support skills also decayed significantly after six months of disuse, suggesting that reliance on AI could weaken practical abilities if not balanced with continued hands on practice.

How AI can be used to support learning

Rather than an all or nothing offload, the episode emphasizes desirable difficulties and retrieval practice. When learners generate answers themselves before using AI as a check or for guided examples, they tend to retain skills more deeply. The guest notes that AI can improve outcomes in some scenarios by providing feedback or helping with practice tasks, but complete offloading may reduce long term mastery.

Novices versus experts

Research suggests that novices may be more susceptible to skill decay when AI reduces active problem solving, whereas experts may retain skills longer though they are not immune. The podcast references studies on mathematics learning with calculators and the idea that fundamental cognitive abilities may be less impacted by practice. The bottom line is that skill and knowledge are highly sensitive to how one uses AI, whereas basic cognitive abilities may be more resilient.

Real world settings and implications

The discussion covers radiology and surgical robotics as areas where AI is increasingly integrated into practice. A key concern is ensuring that clinicians remain capable of functioning without AI and can adapt to updates or failures of AI systems. The podcast also touches on aviation autopilot and the tendency for pilots to rely on automation, reinforcing the need to practice manual control to prevent skill atrophy.

Practical recommendations for workers

McNamara advises participants to tailor AI use to the skill they want to maintain, avoid offloading cognitive work completely, and practice retrieval and generation before seeking AI assistance. She also recommends testing performance without AI to gauge actual competency and interrogating prompts and outputs to prevent overreliance and overconfidence. In radiology, the idea is to maintain human and AI joint cognition by verifying AI findings with independent human review.

Future directions and current research

The guest discusses ongoing collaborations with computer science, medicine, and robotics to examine skill development and maintenance in radiology and surgical robotics. They are planning longer scale studies to assess longer term skill acquisition and cognitive offloading effects, including potential overconfidence and skill decay, and to develop AI integrated study designs that preserve or enhance learning outcomes.

Closing insights

The episode frames AI as both a potential enhancer and a risk, underscoring the importance of balanced usage. It closes with an invitation to revisit these questions as technology evolves and more data accumulate, underscoring the dynamic nature of AI in the workplace and the ongoing need to safeguard core learning and expertise.