top of page
Search

The Power of Perceived Effort PART 1: When polish stops proving the work

4 days ago
9 min read

Take the same content.


Version one is plain text. Clear paragraphs. Well-organised bullet points. The argument is sound, the recommendations are sensible and there are no obvious gaps.


Version two contains exactly the same ideas and conclusions, but now there are coloured callout boxes, visual hierarchy, a few well-chosen icons, punchier headings and perhaps a subtle graphic that brings the core argument together.



For most of us, version two wins almost instinctively. The ideas are identical, but the second version somehow feels more credible, considered and valuable.


For a long time, there was good reason for that instinct. Professional polish was harder to come by. Someone had spent time thinking about the structure. A designer had made decisions about hierarchy, spacing and visual emphasis. Someone had rewritten the headline five times. Someone with fresh eyes had proofread the document and noticed the awkward sentence on page 14 (not to mention the typo on page 1 everyone missed).


In other words, the finished artefact contained visible residue of otherwise invisible effort.

This investment of time, multiskilled team coordination and resources signalled high-quality content with enough importance to justify this effort. We could not directly observe the thinking that went into the work, so we learnt to infer it from the signals we could see.

Almost overnight, AI changed the effort of producing the signals.


The effort heuristic


There is a psychological basis for our tendency to equate effort with quality, called the “effort heuristic. In a series of experiments published in the Journal of Experimental Social Psychology, participants valued creative works more highly when they believed more time and effort had gone into producing them. Importantly, the effect became stronger when the intrinsic quality of the thing being judged was difficult to determine.


This matters enormously in professional services.


If you are buying a car, there are performance specifications, price comparisons, expert and peer reviews - not to mention the physical experience of taking a test drive.


How do you immediately assess the underlying quality of ideas? A strategy, research interpretation, brand idea, commercial recommendation or piece of consulting advice?

Quality reveals itself over time, so we often rely on proxies.


Does the presenting team seem to ‘get it’? Does the presentation appear bespoke and as though somebody cared about bringing the insights to life? Are there any ‘mistakes’ that signal the content may not be suitably reviewed and reliable? Does it look as slick and polished as ‘reference’ content produced by other expensive credible agencies?

Visible effort becomes evidence for invisible thought.


Research on what psychologists call processing fluency adds another layer. Things that are easier for us to process tend to produce more positive responses, while research into website design has found that identical information presented with stronger aesthetic treatment can be judged as more credible.


So polish has traditionally performed two quite different jobs.


First, it has functional value. Good writing makes complicated ideas easier to understand. Good design directs attention, clarifies hierarchy and helps us see relationships in information. Polish has also had signalling value. It tells us, consciously or otherwise, that substantial care and expertise probably sit behind what we are looking at.


Quality signal inflation


Generative AI is extraordinarily good at manufacturing many of the things we historically associated with professional effort - almost instantly.


Scaled synthesis of reports, research and big data for specific insight goals. Structured arguments, with frameworks and concept visualisations. Concept summaries and insight articulation. Clean prose and grammatical consistency. Presentation layouts, image generation and consistent design formatting. Endless options and refinement.


There is growing evidence that it can materially improve professional productivity. In an experiment published in Science, 453 college-educated professionals completed realistic writing tasks. Those given access to ChatGPT completed them around 40% faster, while independent evaluators rated their output around 18% higher in quality.


That is an extraordinary productivity gain but it creates a signalling problem - receiving polished professional writing tells us much less about the effort or expertise behind it.

AI-quality grammar using ‘best practice’ prose for a given application may soon simply be expected – perhaps akin to younger generations raised on electronic music and auto tuned singers feeling that ‘old skool’ music feels a bit loose and rough to them.


The same applies to design with ‘best practice’ principles becoming table stakes. A visually impressive slide once implied that somebody had selected the hierarchy, worked through the information, decided what deserved prominence and translated the argument into a visual form. Increasingly, software can generate many of those surface characteristics before the author has necessarily resolved the thinking underneath them.


We might think of this as signal inflation. We increasingly expect more polish because producing it is easier. When easier for everyone to obtain, its absence can become more conspicuous at exactly the same time as its presence becomes less impressive.


When looking better and being better come apart


For consulting and other knowledge professions, this becomes more than an aesthetic issue and gets to the very core of our CVP and what we supposedly charge for.

A fascinating field experiment involving hundreds of Boston Consulting Group consultants tested the performance of professionals using GPT-4 on realistic consulting tasks. While performing impressively on some tasks, the takeaway is that AI can make an answer clearer, more structured and more convincing without guaranteeing that the underlying judgement is right. That possibility should make anyone who sells ideas slightly uncomfortable.


Imagine two recommendations a client must choose between. The first is using a tired old template but reflects genuine familiarity with the client’s organisation, contradictory customer evidence, commercial constraints and a difficult trade-off that the team has spent weeks debating. The second is elegant, fluent and beautifully structured, but is essentially the statistically plausible synthesis of what a good strategy recommendation normally sounds like.


The temptation is to be at least a little swayed by which one looks more expensive, familiar and easy to consume rather than which one contains evidence of quality judgement.


The AI disclosure paradox


Audiences are adapting.


A 2024 experiment examined how people evaluated writing when they were told AI had been involved in producing it. Disclosure of AI assistance, particularly when AI had helped generate substantive content rather than merely edit it, reduced average quality ratings for both argumentative essays and creative stories.


More recent work has found a similar effect around creative output. Identical work can be evaluated differently depending on whether people believe it was produced by a human, AI or human-AI collaboration, with lower perceived effort helping explain negative reactions to AI-labelled output.


A 2026 study examining workplace and social-media communication goes further, identifying an “AI penalty” in which AI-mediated communication was rated as less trustworthy and authentic than human communication. Intriguingly, participants also believed AI use should be disclosed, creating what the authors describe as a disclosure paradox: people want transparency, but may penalise you when you provide it.

So we arrive at a peculiar contradiction. AI can genuinely improve the output and we increasingly expect professionals to use it and the resulting standard of communication to be higher. But once we suspect that AI produced the polish, we may simultaneously downgrade what that polish tells us about the person who produced it.


Damned if you do, damned if you don’t?


Perhaps. Maybe we are simply entering a transition period in which the old quality signals are weakening before we have collectively learnt what should replace them.


Will mistakes be the new signal for authentic human effort?


This brings us to one of the more provocative possibilities for the future of value assessments.


If flawless prose and immaculate layouts increasingly trigger the thought “AI probably did this”, could a small imperfection make professional communication feel more authentic? Will tomorrow’s signal of human effort be today’s tarnish – a typo, slightly awkward grammar or hastily hand drawn visuals?


Probably not. There is still likely good reason to care about basic errors, at least in our current paradigm. For now, at least, deliberately misspelling strategy on slide nine is unlikely to become a clever authenticity hack. There is, however, a more interesting adjacent idea.


Consumer research has found circumstances in which a minor negative detail can actually improve an otherwise positive evaluation, an effect researchers describe as the “blemishing effect”. The lesson is not that mistakes are good but that perfection is not always synonymous with more credibility.


Emerging signals of value


Traditionally,


Polish → inferred effort → inferred expertise → perceived value


AI breaks one of those links and polish no longer necessarily implies effort in the fundamental way it used to – the old signals are becoming unreliable.


So now,


Abundant polish → uncertain effort → uncertain expertise or distinctiveness


This means the emerging questions, especially for anyone who is in the ‘business of ideas’, are:

  • What will the next signals of value look like in a world of rapid quality signal inflation?  

  • How will ‘solutions’ best be commoditised in a world where confident answers are turning into noise?


Perhaps tomorrow,


Polish + proof of judgement → credibility and value


For professional communication, human ‘imperfection’ may only signal value when it is intellectual rather than grammatical or aesthetic. This might be admitting what you don’t confidently know and why, what more exciting claims are not fully supported and where there is a credible alternative explanation. As exposure desensitises us, and we become less trusting of confident sounding answers, perhaps an increasingly transparent style of human voice will carry more credibility.


That kind of transparency may become increasingly powerful precisely because generative systems make confident completeness so easy to produce. While sloppiness does not signal authenticity, honest fallibility may. It also demonstrates a human thought process.

Perhaps the greatest signal of value in a ubiquitously polished future will be evidence of your judgement. Not just your ability to provide answers but your ability to ask the right questions. Your ability to curate the journey to solution.


In Part Two we dive further into practical tips for communicating ideas (and signalling their value) in the era of AI and how unbundling consulting models may need to evolve.


Authors



Tai Rotem is a consulting partner at MYMAVINS with several decades experience in consumer, financial services, public health, and social research.


Reach out to him at Tai@mymavins.com.au

 






Mitzi Cruz is the creative operations manager at MYMAVINS with seasoned leadership experience managing large-scale web development and design teams. Combining a career in creative direction and operations management with a background in Psychology, she translates research into compelling visual narratives and campaign-ready assets.






Bibliography


 
 
 

Comments

bottom of page