Artificial intelligences are revolutionizing many aspects of our lives. And scientists therefore begin to study how to best integrate these powerful and rapid analysis tools with the work of human beings. At the MIT Center for Collective Intelligence they asked themselves, for example, whether men and women work better together, or alone, and the answer, contained in a study published in Nature Human Behavior, is not exactly what they expected: in many cases In fact, it seems that by collaborating with us the AI only worsens their performance.
The study is a large meta-analysis that has surveyed the research carried out in recent years on the interaction between humans and artificial intelligence. In total, MIT researchers analyzed 160 experiments published in peer-reviewed journals between 2020 and 2023, in which men and women were tested on various types of tasks, but always comparing three possibilities: results obtained by men alone , AI results, and results obtained by men and AI collaborating with each other.
Humans and artificial intelligence are not good at collaborating
On average, the results obtained by humans and artificial intelligences working together were superior to those of humans working on their own, but not than those of artificial intelligences alone. Importantly, studies have not demonstrated the presence of a synergy between humans and machines: the best results obtained by humans and artificial intelligence have always been superior to those that emerge from their collaboration. In other words, for a well-designed AI and for a human being good at a certain task it is better to work alone.
“It is taken for granted that integrating artificial intelligence into a process will always improve performance, but our data shows that this is not true,” explains Michelle Vaccaro, one of the researchers at the MIT Center for Collective Intelligence who carried out the meta-analysis. “There are in fact cases in which it is better to leave some tasks to humans, and others to AI.”
This, at least, when taking into consideration tasks related to decision making, such as medical diagnoses, recognizing deep fakes, or predicting the performance of stock markets. In the case of creative tasks, however, the synergy between humans and artificial intelligence finally seems to be bearing the desired results.
In tasks such as summarizing posts taken from social media, answering questions asked in chat, generating textual content or images, collaboration between men and women achieves better results than either of them can do individually. According to the authors of the study, this stems from the dual nature of such tasks: they require creativity, something in which humans are still superior to machines, but also analytical and repetitive work, which is what AI excels at. To produce an image, for example, you need artistic inspiration – the prerogative of the human being – and then its detailed execution, which is something in which artificial intelligence gives its best.
This is why the authors of the study believe that in the future the way in which artificial intelligence is used in the workflow of companies and organizations will need to be rethought, to get the best from both the programs and the human employees who use them. “Let AI do the research, pattern recognition, prediction, and data analysis, instead leveraging human abilities to notice subtleties and nuances, and understand context,” concludes Thomas Malone, co-author of the study. “The more we explore the potential of these collaborations, the clearer it becomes that the future is not about replacing humans with AI, but rather finding innovative ways to make these two elements work together effectively.”