MIT study: ChatGPT increases productivity, has potential to close the inequality gap
The big question with generative AI these days is whether tools like ChatGPT will widen the inequality gap or empower workers with newfound skills and abilities.
A study(opens in a new tab) from MIT’s Department of Economics designed to answer this question found that participants using OpenAI’s ChatGPT increased productivity and the likelihood that they would use ChatGPT in future tasks. In the controlled study, this implies that “the technology will be more strongly complementary to human workers,” meaning it favors tools like ChatGPT as a way to empower workers. But how these tools are actually implemented in the real world remains uncertain.
Unlike earlier AI tools which raised concerns about automation of “routine” tasks, deep learning tools like ChatGPT are capable of executing more complex, creative tasks like writing and design. How generative AI is implemented in the workforce could negatively or positively impact labor inequality. “Inequalities between workers could either decrease if lower-ability workers are supported more by ChatGPT or increase if higher-ability workers have the skills necessary to take advantage of the new technology,” said the study.
The experiment comprised 453 college-educated professional and randomly assigned half of the participants with ChatGPT after completing their first assignment. The assignments were writing-based tasks including press releases, short reports, and “delicate emails,” mimicking those that grant writers, marketers, consultants, data analysts, and HR professionals would do in their day-to-day work.
The study found the group that was given access to ChatGPT decreased in time taken to accomplish a task by 11 minutes and increased in quality. Notably, the performance of the treatment group (those using ChatGPT) increased between their first assignment (without ChatGPT) and subsequent assignments (with ChatGPT), which the study concluded could close the inequality gap between skilled and unskilled labor.
This has been anecdotally true for anyone using ChatGPT. But the study provides hard evidence that workers armed with ChatGPT can be more productive and perform tasks better. Yet, how this plays out in the real world remains to be seen. Is this proof that ChatGPT should be taken as a new tool in workers’ toolkits? Or will companies interpret this as evidence that generative AI can successfully replace entire jobs? Ultimately, this study underscores how the implementation of generative AI depends on a wildly complex and unpredictable factor: human nature.
