Â鶹ÒùÔº

How artificial delegates can help us act more socially—yet still fail to achieve collective goals

Can artificial delegates—autonomous agents that make decisions on our behalf—help us reach better outcomes in situations where collective failure looms, such as climate change policymaking or the urgent response required during pandemics?

A new behavioral experiment led by Professor Tom Lenaerts (VUB/ULB) sheds light on this pressing question. The findings are surprising: individuals who entrust their decisions to digital representatives tend to behave more generically pro-socially, but this does not automatically lead to better outcomes. The study is in the journal Proceedings of the National Academy of Sciences.

The study demonstrates that people who delegate their choices to artificial agents contribute more to the collective good, even when they have previously experienced negative outcomes or inequality.

"What we observe is that, once people take a step back and allow a configurable AI to decide for them in a social dilemma, their contributions tend to be greater than when they make the decisions themselves," explains Professor Tom Lenaerts. "They appear more willing to do their part—an interesting behavioral shift."

Yet, this increased willingness does not translate into more successful group outcomes. Artificial tools are consistent but lack the human ability to adapt flexibly to the actions of others in the setting. "They struggle to respond effectively to unexpected circumstances or last-minute decisions. They are well-programmed, but not perfect," Lenaerts notes.

Although the use of digital representatives encourages more positive behavior, this alone is not sufficient to tackle complex group challenges effectively. The key lies in sustained human engagement in programming these representatives.

"Delegating to these tools only works if people continue to think actively about how they are designed and configured," Lenaerts emphasizes. "The notion that we can simply hand everything over to technology without investing ourselves is a dangerous illusion."

The findings reveal a crucial dilemma in the evolution towards increasingly digital decision-making: technology can amplify social intentions but does not guarantee collective success. In fields such as climate policy or public health, this could have far-reaching implications.

"Artificial agents certainly have potential, but they are not a magical solution," Lenaerts concludes. "If we want them to truly work in practice, we as humans must continue to take responsibility—not only in the decisions themselves but also in building the systems that make those decisions."

The findings are part of the doctoral research conducted by Dr. Ines Terrucha, in collaboration with both national and international researchers.