AI And Work Will It Bridge Or Widen Skill Gaps

The question of who benefits most from generative AI in the workplace is a significant one. Does it uplift less experienced workers, or does it further enhance the capabilities of high performers? While much research suggests AI acts as an equalizer, some studies have indicated it might favor those already at the top.

A Key Study Discredited Shifting Perceptions

Recently, a widely publicized study supporting the idea that AI helps high performers more has been discredited. This paper, which media outlets and even publications like Charter covered, claimed a genAI tool for discovering new materials significantly boosted high-performing scientists over their lower-performing colleagues.

The author of this research is no longer with MIT, and the university has publicly stated it "has no confidence in the veracity of the research contained in the paper." This development was a shock, as the study had been influential in understanding AI's potential impact on work and inequality. Given its retraction, it's crucial to re-examine what other current research indicates about AI's role in widening or narrowing performance gaps between workers.

Why Understanding AIs Impact on Performance Gaps Matters

This question is vital for several reasons. As Rembrand Koning, an associate professor at Harvard Business School and co-author of a paper on this topic, explains, it helps us understand whether AI will exacerbate inequality. More than that, it has profound implications for less experienced workers. "If [AI] helps people with more experience, [who] tend to be the people who can do things better, it might take out the lowest rungs of the labor market because we think AI can have such large effects," Koning warns. He adds that long-term, "there may be policy solutions or organizational changes that can get around some of these problems…we need to know the facts to start understanding how the economy might respond."

The Dominant View AI as an Equalizer

Most studies investigating this issue find that generative AI tends to have an equalizing effect. It appears to help less experienced, lower-performing workers more than their highly experienced, top-performing peers. With the material sciences paper now retracted, very few studies contradict this general finding.

This "AI-as-an-equalizer" outcome has been observed across various types of work, from customer-support roles to tasks involving writing and consulting. Two important aspects of these studies are worth noting when considering their applicability elsewhere. First, the tasks examined are areas where AI is already known to perform very well. Second, in these studies, the output from the genAI tool is often very close to—or sometimes is—the final product. Therefore, the mechanism by which AI levels the playing field in these instances is largely by performing a substantial portion, if not most, of the actual work.

When AI Might Widen Performance Gaps

Despite the prevailing trend, a few papers suggest that AI can widen performance gaps. These include research on college students in a debate competition and a study of entrepreneurs in Kenya. Both studies underscore that human judgment is a crucial factor in determining who benefits from AI.

In the Kenya study, for example, an AI tool offered entrepreneurs a variety of advice. It was then up to the individuals to decide how to use this information. Harvard Business School’s Koning, one of that study's co-authors, explains, "Those who have the judgment do better because they’re able to be like, ‘Yeah, these 10 pieces of advice—six are terrible, two are neutral, two are actually really good. Let me focus on the two that can actually help my business.’"

So Will AI Widen or Narrow Performance Gaps

The current perspective suggests that AI's impact on performance gaps depends significantly on the level of autonomy within a job. If your role involves setting your own goals, deciding how to achieve them, and making numerous decisions daily, AI is more likely to benefit you if you are already a high performer with strong judgment. Conversely, if you are in a more constrained role that requires less decision-making and involves a predefined set of tasks that AI can perform well, then AI is more likely to narrow the performance gap.

Ben Weidmann, director of research at the Harvard Skills Lab and co-author of a related paper, offers a way to think about this: "When you start your day and you’re looking at the possible number of paths you can go down…how many paths could I possibly choose? And then how different are the end destinations of those paths?" He hypothesizes, in line with this view, "that jobs where there are more paths are going to disproportionately have the AI gains go to people who are high-skilled" in their decision-making abilities.

Koning provides another useful distinction: "The other way to think about it is [whether] the AI [is] producing the inputs or the outputs for your job." In many studies where AI narrows performance gaps, the AI’s output effectively becomes the worker’s output, perhaps with some editing. Koning highlights that in the debate and entrepreneurship studies where gaps widened, the AI’s answers were the inputs for complex tasks like debating or making critical business decisions and implementing them.

This analysis offers a condensed look at a complex question. For those interested in a more detailed exploration, including further implications for different job types, additional analysis is available.