Working through a giant social experiment

Columnist Louise McLaren reflects on the ease and cognitive surrender of it all.

illustration of human head offloading shapes

I’ve recently become more conscious of a friction inside my brain: the uncomfortable rubbing together of two opposing inclinations. On the one hand, use AI to create a shortcut and save labour: marvel at the quality of its outputs, delight at the time-saving implications, bash something out and rush off to the next thing. 

On the other hand, feel that dopamine rush from the shortcut to be tempered with queasy discomfort. Because I know, if I let my addled brain pause for a moment, that various risks and downsides may quickly come back to bite me in the short- and mid-term. 

Are we over-promising, or selling ourselves short?

I might propose outcomes from a project that I simply cannot deliver, because the timings are insane, or the data will never be able to answer what we’ve blithely promised. I might write a debrief that has errors in it – that overstates certain insights, and which I can’t back up when I’m exposed, presenting to the client.

It will probably be very evident to the client where AI has been used (because let’s face it, it often is) and they might question what they’re actually buying. And I might erode the differentiation I and my team bring as human beings, writing pitches and debriefs that look like those of any other agency.

All of these are real risks, and they’re very much on my mind these days. They always have been, and they’ve informed the cautious adoption I’ve taken towards AI tools. But those risks weigh on my mind more now – because AI has got better. 

It’s astonishing how good what Claude can produce can be, within minutes. And it’s breathtaking how quickly this change is happening to me, to us. The quality of the AI output heightens the seductive lure of the shortcut. 

I happen to be fortunate enough right now to value some shortcuts: we’re busy. And not only are we busy with projects and proposals, but deadlines are tight. They seem to be getting tighter. When the day is a mad juggle, it’s human nature to find the paths of least resistance. 

“It’s breathtaking how quickly this change is happening to me, to us. The quality of the AI output heightens the seductive lure of the shortcut.”


I remain unconvinced by the common narrative I’ve seen that valorises the quality of the prompts we write; like real effort has to go into this, cognitively, practically, and that this proves to us that we’ve really applied ourselves, the work has our own unique stamp on it. Rubbish. For a long time, and on many tasks, the LLMs I use have spat out remarkably useful stuff in response to a prompt it’s taken me mere seconds and absolutely zero reflection to write.

In the context of this seductive lure of the shortcut, it helps that LLMs and the other mainstay AI tools in research don’t always get it right. Whether it’s my attempt to buy  the perfect winter jacket for my son after painful searching myself – Claude gave me a bunch of ideas that were not on-brief or not available – or the analysis on an AI qualitative research platform that overstates certain findings, or a data analysis tool that claims to be drawing from the total sample but is not. It takes an active mental effort to hold on to these failures, and remember they can crop up at any point, and more subtly.

From offloading to cognitive surrender

The problem runs deeper than this, though, and it’s a longer-term and more profound one. The stronger the pull to take the easy route, the more we risk weakening our critical thinking muscles. This is not a revelation. We’ve been talking about this risk – in this sector, in society – for some time.

But now we know more about this. It’s not just cognitive offloading (outsourcing a bit of thinking in the moment) but cognitive surrender – giving ourselves over in a more unguarded way to AI. Of course, this is a fast-moving issue: there is a limit to what evidence we have as yet for how AI is affecting our brains.

A recent study showed that using LLMs as opposed to classic search engines for a set task led to shallower knowledge acquisition1.  Another paper published this year, based on a series of randomised controlled trials (a gold-standard methodology) showed that AI assistance reduces persistence at a task and impairs the ability to work without it, once it is removed2.

On the other hand, there are early indications that using AI as a collaboration tool that provides feedback can enhance learning (from a small-scale exploratory study)3 , or using AI outputs as examples to be emulated can benefit future performance4, even in cases when the AI is no longer available. It’s understandable that there are benefits to be found. For those looking to exlore this issue further, this new resource from MIT and academics across a range of institutions to track the impacts of AI in a broader sense is fascinating5.

Overall, it feels worth just taking a beat. We don’t yet know how AI will affect how we think, and the skills that we apply to our work. We should recognise that we’re going through a giant social experiment. 

I keep coming back to a point made by Amy Kean that I raised in a discussion about AI and ethics at TMRE in September. This is one for employers: given what we’re learning about how AI could blunt our thinking skills, it starts to possibly become a moral issue, Amy said, to push our staff to use AI – we may not simply be affecting their jobs now, we may be affecting their brains longer term.

This may sound dramatic, but the point is – we just don’t know.  And we need to retain our critical thinking abilities in order to know.

Louise McLaren is managing director (London) at Lovebrands and a columnist for Research Live. You can read all of her columns here.

1: https://pubmed.ncbi.nlm.nih.gov/41163786/

2: https://arxiv.org/pdf/2604.04721

3: https://link.springer.com/article/10.1007/s40593-025-00521-6

4: https://arxiv.org/pdf/2502.02880

5: https://impactbench.media.mit.edu/about

We hope you enjoyed this article.
Research Live is published by MRS.

The Market Research Society (MRS) exists to promote and protect the research sector, showcasing how research delivers impact for businesses and government.

Members of MRS enjoy many benefits including tailoured policy guidance, discounts on training and conferences, and access to member-only content.

For example, there's an archive of winning case studies from over a decade of MRS Awards.

Find out more about the benefits of joining MRS here.

0 Comments


Display name

Email

Join the discussion

Newsletter
Stay connected with the latest insights and trends...
Sign Up
Latest From MRS

Our latest training courses

Our new 2025 training programme is now launched as part of the development offered within the MRS Global Insight Academy

See all training

Specialist conferences

Our one-day conferences cover topics including CX and UX, Semiotics, B2B, Finance, AI and Leaders' Forums.

See all conferences

MRS reports on AI

MRS has published a three-part series on how generative AI is impacting the research sector, including synthetic respondents and challenges to adoption.

See the reports

The only place...
to become an AI expert

Training

Conferences

Reports

ProgrammesĀ 

Ā