The Blurring Boundaries of Behaviours

AI is changing our behaviour. It is impacting on the information-seeking behaviours of our customers. This is impacting on the services we are delivering, but how? What have you noticed?

At the recent AI Knowledge Café, people reflected that now the first step in many actions is to ask AI a question. Another admitted they had started doing the opposite, deliberately writing first and only using AI afterwards to challenge thinking. Whether a sceptic or an adopter, it is something we are slowly beginning to rely on.

As healthcare librarians, we are often interested in what people know, how they learn, and how they make decisions. Increasingly, AI forces us to add another question: what happens when some of those activities are delegated?

From searching to asking

Information-seeking behaviour is evolving. Instead of constructing a search strategy, selecting databases, refining keywords and evaluating results, many people now turn to AI. In a fast-paced world, there is a need for immediate synthesis rather than a list of sources. Convenience and speed are powerful attractions.

Few clinicians have spare hours to spend refining searches between clinics, ward rounds and administrative work. Yet the very convenience that makes AI attractive may also alter the habits that underpin evidence-based practice.

A summary feels like understanding. Sometimes it is. Sometimes it is merely the appearance of understanding.

The “use it or lose it” question

The phrase that surfaced repeatedly during the café was “use it or lose it”. Participants wondered whether repeated cognitive offloading might gradually weaken skills such as writing, searching, appraisal and analysis. Many recognised changes in their own behaviour.

As librarians, this concern feels familiar.

For years we have encouraged users to move beyond Google and develop critical appraisal skills. Now we are beginning to ask another question: if AI performs more of the thinking process, what skills should people continue to practise for themselves?

Learning often happens during the effort of selecting information, organising ideas and constructing an argument. Remove all that effort and something valuable may be lost.

When confidence looks like expertise

Another theme was trust.

AI can exhibit perceived warmth, confidence and affirmation. It can make interactions feel almost human. A convincing answer remains capable of being wrong.

This matters because humans are naturally inclined to respond socially. We thank AI systems. We apologise to them. Some users describe them as companions, coaches or partners.

One participant observed that people increasingly say, “Let’s see what Copilot thinks.” A seemingly harmless phrase, but perhaps one that subtly transfers responsibility from person to machine.

Evidence pollution and the tsunami of content

The discussion became particularly animated when we explored evidence quality.

Library staff shared examples of fabricated articles, references and digital object identifiers (DOIs) appearing in AI-generated outputs. Others described encountering websites, courses and resources that contained inaccuracies despite their professional appearance.

The evidence is becoming polluted. If synthetic summaries, citations and publications are continually reused and incorporated into future content, distinguishing genuine evidence from generated material may become increasingly difficult.

In this environment, the traditional librarian skills of verification, appraisal and source checking become essential.

Accountability

Accountability remains with healthcare professionals. Recently NHS Resolution released the criteria of clinical reliance claims involving AI:

• Did the clinician appropriately rely on the AI, or fail to apply their own judgment?
• Was the AI used in accordance with guidance, training, and regulatory approval?
• Were any known limitations / performance concerns of the AI system disregarded?

We use the phrase “human in the loop” to explain this accountability. Meaningful oversight needs to go beyond merely clicking an approval button. True oversight requires expertise, time, authority and the willingness to disagree with the technology when necessary.

The changing role of librarians

Perhaps the most encouraging aspect of the conversation was the recognition that Knowledge and Library Services have a significant role to play in this emerging landscape.

The AI age does not eliminate the need for librarians. If anything, it highlights our expertise in different ways.

Participants identified growing needs around AI literacy, prompt design, evidence verification, copyright, governance and evaluation of AI-generated content. Libraries may increasingly help users navigate not just information abundance, but information uncertainty.

A final reflection

AI works best as a scaffold, critic, editor, checklist or source of alternative perspectives. It is less effective as a substitute for judgement, reflection or accountability. We need to establish what boundaries need to be drawn for society to thrive.

Susan Smith, Emily Hill  Mid Cheshire Hospitals NHS Foundation Trust