AI and Humans: A Complex Partnership in the Workplace

The rise of artificial intelligence agents — bots that not only provide information to users but also execute tasks for them — has sparked talk of an agentic staff working alongside humans and managing other bots.

However, a Massachusetts Institute of Technology research paper, “When combinations of humans and AI are useful: A systematic review and meta-analysis,” found that human-AI collaboration, on average, “performed significantly worse than the best of humans or AI alone.”

“This was our most surprising finding,” co-author and MIT Sloan professor Thomas Malone said in an MIT article. “Some of the most important and interesting use cases for AI involve a combination of humans and computers. Many people would have assumed the combination would be quite a bit better, but it was statistically significantly worse.”

Possible reasons for this unexpected finding include the presence of communication barriers, trust issues, ethical concerns, and lack of effective coordination between humans and AI systems, according to the paper, which was published in the journal Nature Human Behaviour.

How people viewed AI — whether as a magic tool or simply a machine — also mattered.

“For example, people often rely too much on AI systems…, using its suggestions as strong guidelines without seeking and processing more information,” the paper said. “Other times, however, humans rely too little on AI…, ignoring its suggestions because of adverse attitudes towards automation.”

If an AI system wrongly identified a tumor on an X-ray as benign, a doctor could accept the diagnosis without challenging it, for instance. On the other hand, a financial analyst could ignore a prediction by the AI if it had made a mistake before — even if the AI outperforms the analyst overall.

The paper also measured human-AI performance when it came to two tasks: creating content and making decisions.

It turns out that human-AI collaborations work well in content creation. While creative skills are involved, there’s also routine in the task, like filling out an image or predicting the next word in a sentence, which AI is good at doing, according to the paper.

“[G]enerating many kinds of text documents often requires knowledge or insight that humans have and computers do not, but it also often requires filling in boilerplate or routine parts of the text as well,” the paper said.

When it comes to decision making, however, human-AI synergy led to “significantly negative” performance, according to the paper.

Human-AI synergy often faltered because it didn’t tap complementary strengths effectively, the paper found. AI can process large datasets quickly, and humans are better at contextual interpretation and ethical judgment.

For better results, the AI system could have been given “only the parts of the task for which they were clearly better than humans” and vice versa.

The difference between human-AI synergy and augmentation came down to the benchmark used in the paper, which reviewed more than 100 experimental studies over three years.

The authors defined human-AI augmentation as using AI to improve human performance on a task. The AI might provide suggestions, automate routine work, or do other things to enhance human capabilities. The researchers measured success based on whether the combined system performed better than a human working alone. It didn’t matter if the AI alone did better.

Human-AI synergy, on the other hand, sets a higher bar. It occurs when the combined efforts of the human and the AI lead to better performance than either the human or the AI could achieve alone. In this scenario, the human and the AI bring complementary strengths to the task, resulting in a truly collaborative outcome that exceeds the sum of its parts.

For example, consider the task of designing marketing content. A human might excel at understanding the target audience’s emotional triggers, while AI can rapidly generate multiple variations of visuals and text. If the combined system produces more engaging content than the human worker or the AI could on their own, that’s synergy.

Augmentation focuses on helping humans perform better while synergy aims for a level of performance neither could reach individually.

“[B]ehind every AI success story lies human effort and ingenuity,” according to a Salesforce blog post unrelated to the paper. “While AI tools can augment — and sometimes even replace — certain tasks, the real magic happens with human guidance. It’s people who train these systems, collaborate with them, interpret their outputs, and ultimately make the final decisions. [Workplaces] still rely on human judgment, creativity and the unique perspectives only we can bring.”

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

The post AI and Humans: A Complex Partnership in the Workplace appeared first on PYMNTS.com.