Fast Food vs. Slow Science: Traveling at the Speed of Our Souls with AI

The old way: leaving the sould behind Vs The Deep AI Way: Conversing with the data

We are currently witnessing a massive backlash against using genartiveAI in qualitative research. Calls for "Slow Scholarship" and "Handmade Research" are growing louder, positioning themselves as the moral antithesis to the encroaching wave of artificial intelligence.

Recently, I analysed the language used in these critique and I found a rhetorical trap that is becoming all too common. The debate is being framed as a stark, reductive binary: you are either a thoughtful, organic artisan, or you are a mindless, industrial machine.

But this framing is wrong. And worse, it prevents us from seeing the Third Way: a mode of research where technology supports depth rather than destroys it.

The Rhetorical Trap: Fast Food vs. Organic Gardens

When critics describe AI-supported analysis, they almost exclusively use a "Fast Food" or "Machine" frame. The vocabulary is telling:

Adjectives: Faster, efficient, hurried, standardized, superficial, artificial, inauthentic.

Verbs: Generate, run, convert, feed on, process.

These words strip the research of intellectual weight, implying that using AI is an act of mechanical production—metrics over meaning, speed over substance.

In contrast, the "Handmade" approach is wrapped in an "Artisan" or "Gardener" frame:

Adjectives: Organic, slow, intimate, healthful, flavourful, ambiguous, imperfect, whimsical.

Verbs: Nurture, root, embrace, stimulate, grow.

It’s a powerful narrative. Who wouldn’t prefer a "flavourful, organic" insight over a "standardized, bad sauce"? But this binary equates the use of digital tools with a lack of soul. It assumes that if you use AI, you have stopped thinking.

The Literacy Gap

I do not deny that there is a surplus of low-quality work currently flooding the field. We see computer scientists attempting to tell qualitative researchers how to "automate" analysis in the name of speed—a goal most qualitative researchers do not share.

This perception is further fuelled by qualitative researchers who lack the necessary AI literacy to use the technology properly. We frequently see colleagues wasting weeks conducting studies based on deeply flawed prompting strategies—such as forcing comparisons between manual coding and basic AI outputs. These are methodological dead-ends that an AI-literate practitioner could have identified as futile from the very start.

There is still a lack of knowledge out there, and this is understandable, as the technology is new. Instead of rushing to condemn it, maybe we should take the time (slowing down the hype) to learn how to use it properly instead of rushing to publish articles pointing fingers. A lot of reviewers and editors are not yet capable of separating the wheat from the chaff.

True AI literacy requires knowing how LLMs work and how you need to interact with them. I bet that many critics don't know this—they are just using ChatGPT in the "quick and dirty" way. One can criticize that approach, but one cannot generalize it to claim that all AI-supported analysis lacks intellectual weight and thinking.

Just as a quick test, do you know about:

How an LLM processes data (not only how it produces output)

Context windows, input/output limits
Temperature
System prompts
The difference between conventional and reasoning models
The non-deterministic nature of LLMs and how to make use of it in a positive sense
RAG (retrieval augmented generation)
Sycophancy and what to do about it
The differences, also in terms of privacy, between using a general purpose chatbot and calling on an LLM via an API connection

When you understand these concepts, then you also need to understand that you are not asking an LLM for answers (this is what most people still do).

Developing Questions, Not Answers

We need to use GenAI to develop questions we can ask about our data, not to generate answers. This is what got me excited in the first place when I realised this potential.

We can now interact with our data more or less directly. We can ask questions about it—without the laborious process of coding. The part of being able to interact directly with your data also applies if you use non-coding approaches.

I recognize that some colleagues view manual coding as an intrinsic part of the craft, a sentiment I understand after thirty years in the field. However, we must distinguish between the intellectual spark of building a coding frame—which is indeed engaging—and the clerical repetition of applying it to endless transcripts. The real excitement returns only when working with the coded data.

With the proper use of GenAI, we can skip the repetitive middle and engage with that 'interesting part' from the very start. But to be clear: this does not mean analysis is done in two minutes. That is a dangerous illusion. Because GenAI is designed to always provide an answer, unskilled users often mistake a quick output for a valid result. They treat the tool like a vending machine, and then complain about the 'black box' when they cannot trace how the LLM arrived at its conclusions. The result is either publication with questionable quality or a dismissal of the technology entirely, simply because the researcher failed to interrogate the machine.

Travelling at the Speed of Our Souls

I want to close with a famous parable, often retold in various forms (sometimes it is South American explorers and indigenous porters, sometimes it is about the Bedouin and the camel).

A group of Western explorers is rushing through the jungle with local porters. They push hard, day after day, to reach their destination. Suddenly, the porters stop and sit down, refusing to move. The explorers grow frustrated and ask, "Why have you stopped? We are not tired!"

The head porter replies: "We are not tired. But we have moved so fast that our souls have been left behind. We must wait here for them to catch up."

Critics assume that using AI is the act of rushing through the jungle—that we arrive at the destination (publication) without having experienced the journey.

But in my experience, deep AI usage is actually the act of sitting down with the porters. When I use GenAI properly, I am not speeding past my data. I am stopping to converse with it. I am using the tool to dig deeper than I could alone, to find the "flavor" and "ambiguity" that the slow movement prizes.

We don't need to choose between "fast food" and "starvation." We can cook with modern tools, as long as we remain the chef. The goal is a workflow where our qualitative souls are never left behind.

Try it for yourself and register for a free trial for QInsights.