Nov 18, 2026
5 Views
0 0

What is a researcher’s worth?

Written by

Why the debate over research isn’t a technology story – and why effortless AI makes human friction our scarcest asset.

In a watercolor aesthetic, a boat with various scientific objects on it in the middle of a body of water. Where will they all go?
Image created by author, with Midjourney.

We’re no stranger to this question.

For many years, research practitioners have talked with one another about how to “mature” teams so that they understand the value of our work in bringing success to their ventures. Or when we share the results of our research, we are peppered with questions about whether the findings are real, whether the methods are too weak or lack representativeness, or whether our subjects are even telling the truth when they express to us what they need.

But it’s not just research practitioners. Researchers in academia have been stripped of the funding they need to advance science. Academic departments have shuttered, and scholars’ voices have been dismissed as part of major civic scientific efforts, like the Census, the Environmental Protection Agency, or the Department of Education.

Put in this context, contemporary discourse around artificial intelligence replacing our jobs feels less like “the reason” than one reason among many when people question whether researchers (and research, in general) matter.

Over the last ten years, I’ve worked as an industry researcher while keeping a foot in academia, and have had some of the best work experiences of my life. There is a joy and a magic I find in what happens when you bring experts of different disciplines together – research, business, engineering, design, and more – to solve complex puzzles.

So if that cross-disciplinary magic is so real — and so productive — why do researchers in both companies and public life spend so much of their careers on the defensive? Most hot takes right now answer that question with new kinds of work: prompt faster, automate your synthesis, prove your return on investment.

As a social scientist, I think those takes miss what is actually happening. The debate over a researcher’s worth isn’t really a technology story. It’s a story about how institutions handle inconvenient truths.

Why institutions put research on trial.

Back in 2020, I wrote about one of my favorite scenes from the all-too-accurate show, Silicon Valley. In the episode, technologists watch through a one-way mirror as focus group participants give honest feedback about their app. One participant explained that the app made them feel “totally freaked out,” and another explained that it “made me feel stupid.”

Behind the mirror, the exec got angry. “Clearly none of these people get it, maybe this is a bad group…everyone that I showed the beta to loved it.” His colleague explained that this was the fifth focus group and they all shared the same feedback. She then realized he only gave the beta to his engineer friends. “I wanted to give it to people who would understand what I’m trying to do so I could get useful feedback,” he said. “And with all due respect, I gave it to you — the one person without a computing background – and you said it felt engineered.”

Scene from the television show, Silicon Valley, where focus group participants weigh in on a feature by Pied Piper. User feedback is honest, and critical. Behind a one way mirror, the executive fumes.
Focus group scene from Silicon Valley (S3E9).

Those of us who love the craft of research love talking about research methods. But I find that in many cases, debates about methods in settings like these are not about finding what’s valuable. Methodological nitpicking is a way to avoid disruptions to a preferred plan. In this way, it is a shield against uncertainty – ironically, the very thing that research tries to disambiguate.

Social scientists argue that when under pressure, organizations gravitate toward commensuration – a process by which people working together flatten a messy, pluralistic set of realities into a single spreadsheet metric. The truth is, we do live in a world with a lot of unknowns, and the “winners” of today’s marketplace may not be the same in a few months or years. Under that uncertainty, it is tempting to flatten complex human behavior into click metrics – and flatten researchers into headcount cost compared against a chatbot – mistaking a tidy spreadsheet for certainty.

What AI actually changed: a mirror that rarely talks back.

In 2020, the lead technologist in Silicon Valley had to walk through the door and argue with human beings using their products. In 2026, they don’t have to do that anymore. They can swap the humans behind the glass for a synthetic AI persona trained to give feedback in specific ways.

I’m not here to say that AI is bad for science. I actually love a lot of the applications of artificial intelligence I get to use. I’m more troubled by the discourse around artificial intelligence and research in industry – how people talk about it – and the way it is compared to the human worth of social scientists.

For one, when organizations treat research as theater to rubber-stamp decisions they have already made, synthetic users and LLM summaries offer “good enough” reassurance at a fraction of the cost. Replacing hard-to-reach or marginalized communities with LLM agents doesn’t actually include them, either. It builds statistical caricatures out of dominant internet texts while erasing human consent and real-world inequities – the kind of blind spot that lets an automated camera “touch-up” feature quietly lighten darker skin tones if real people aren’t in the room.

And think about how culture actually moves. A few years ago, if you asked a model trained on historical averages what romance audiences wanted, it would tell you to center a man and a woman falling in love – it never would have bet on a breakout love story about two men playing hockey. Models fed past averages and synthetic data lose the tails of the distribution, the emerging subcultures and edge cases where the future actually starts. Meanwhile, algorithmic platforms mistake digital traces for human intent: a frustrated user clicking three times to fix a broken output looks like “high engagement” in behavioral logs.

Minimalist bell-curve diagram on cream paper: the muted gray dot-grid center is labeled “Historical Averages & Synthetic Personas (‘Good enough’ consensus),” while both highlighted coral outer tails point to “Where Culture & Real Risk Live (Emerging subcultures, global contexts, blind spots).”
Models trained on historical averages capture polite consensus in the middle while losing the tails where new culture and real-world blind spots actually start. Diagram by the author, using Gemini.

Artificial intelligence is an interesting tool, not a good social scientist. But today, AI is giving anxious institutions the frictionless validation they secretly always wanted.

Looking in our own mirror.

Listen, I don’t think researchers are perfect by any means. We’re human after all! I actually think that researchers both in industry and academia accidentally made our worth easy to caricature.

Often professions survive or shrink based on what they claim jurisdiction over – what we diagnose problems to be, to what we attribute their causes, and how we offer treatment. In 2020, I warned that industry research was used too tactically (“go run a focus group”) as a way to drive impact. But over the last decade, much of industry research leaned into method gatekeeping – measuring our worth by how many people we interviewed or surveyed, or how many “insights” we delivered, or how many slide decks were shared.

At the same time, in academia, career incentives rewarded the world’s smartest experts for speaking mainly to peers behind journal paywalls rather than embedding social science into civic and institutional design. For example, a senior professor in my field asked me how I’d like the discipline to advance. I said that we should consider hiring clinical professors to more genuinely bring our findings to the field. He actually laughed at me. “Save it for social work,” he said.

I’m of the opinion that neither of these approaches has really helped anyone. In the former, we have traded using appropriate and defensible methods, asking big questions, and discussing complicated answers for a “study complete” checkbox. In the latter, we have siloed the magic of expert discovery almost entirely to the few people who read a published paper.

Where the magic of research lives.

Why does bringing research, engineering, design, and business together feel like magic? David Stark showed that breakthrough innovation happens in what he calls “heterarchies” – teams that hold competing accounts of worth in creative tension rather than collapsing everything into shipping speed. When AI coding tools let teams spin up twenty working prototypes in a week, velocity is no longer the scarce resource. Knowing which human dream is worth building becomes the scarcest asset in the room. For example, when do you pause a team buried in fifty usability bug fixes to instead ask whether we’re focused on the right value?

Some unfortunate news for my colleagues: humans don’t follow linear product models. They improvise inside their own social groups and cultural ecosystems. They react and adapt to major changes in their lives.

Engineers see systems; product managers see markets and funnels; and social scientists see social structure.

Editorial social network graph on cream paper: three dense slate-gray clusters of interconnected nodes are labeled “Intimate Pockets (Tight-knit peer groups where shared norms form)”; three warm coral hub nodes bridging the gaps between clusters are labeled “Moderators & Bridges (Key connectors holding the wider ecosystem together)”; and a small dashed-line cluster on the far right with ink sparks is labeled “Emerging Subcultures (New scenes taking shape at the edges).”
What social structure actually looks like underneath platform metrics: tight-knit intimate pockets, central moderators and bridges holding the wider ecosystem together, and emerging subcultures taking shape at the edges. Diagram by the author, using Gemini.

Social scientists can find and explain why youth culture over the last twenty years has shifted from public broadcasting to more intimate pockets of the internet. Or why a certain design aesthetic feels exciting in one region and troubling in another, shaped by the cultural meanings people there share. Or how chaotic-looking livestreams with tens of thousands of people are actually held together by the patterned, organized activities of moderators who keep conversation focused and fun.

In The Tacit Dimension, Hungarian-British polymath Michael Polanyi wrote that “we can know more than we can tell.” A vast amount of human knowledge is nonverbal. It includes how we think about cultural norms that we can’t summarize in a survey, or sometimes even in the context of an interview. My bet is that unguided LLMs can’t uncover that tacit knowledge on their own. But social scientists can. We triangulate methods and hold magnifying glasses up to a social problem in a variety of ways, including by testing our assumptions against an accumulated body of scholarly work in purposefully specific subfields.

Worth is built together.

Whether in a product sprint, a university lab, or a public agency, a researcher’s worth was never the focus group they run, how big or small their sample size is, or the paper they publish. Our worth is how we apply our training and our lens to the messy, plural, inconvenient realities of human life, even when institutions are tempted by easy mirrors that boil experience down to a lone metric.

To my research peers: bring your expertise and critical thinking to the work that you do. Show value to your teams by how you and your work “make the strange familiar, and the familiar strange” in the pursuit of a shared understanding grounded in human experience.

To my academic colleagues: get out of the ivory tower. Meet people who do research and engage in practice for purposes that have nothing to do with publishing. Learn new ways to talk about your work that shed both academic-ese and didactic habits in favor of two-way learning.

To my partners in research: resist the seduction of a mirror that always agrees with you. When AI makes spinning up twenty new ideas effortless, speed stops being your competitive edge – knowing what is actually true about people outside your building is. Invite the friction of someone trained to tell you when your favorite plan doesn’t match how humans really live.

We’ve spent years talking about how to “mature” organizations into valuing research, as if worth were a favor. It isn’t. Worth is built together.

Dr. Matt Rafalow is a senior staff researcher at YouTube, and a visiting scholar at University of Southern California. The views expressed in this article reflects the individual opinions of the author.


What is a researcher’s worth? was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.

Article Categories:
Technology

Leave a Comment