The enterprise technology world has discussed at length bias found in AI products, in the data that fuels applications and in the outcomes AI can produce. But one oft-overlooked area is how humans can have bias toward AI insight.
When AI produces suggestions, executives making top business decisions can have different responses, exhibiting bias toward the results, according to new research published in The MIT Sloan Management Review. Co-authors Philip Meissner and Christoph Keding presented 140 U.S. senior executives with a hypothetical business decision, and were told to take an AI-based system's recommendation into consideration.
After measuring executive responses, the co-authors found responses to AI recommendations varied between execs, leading to different decisions — in this case, whether to invest in a new technology and how much.
"We make entirely different choices based on identical AI input," said Meissner, professor of strategy and decision-making at ESCP Business School in Berlin. "This really highlights, as we call it in the title, the human dimension in AI-based-decision making, because we have to acknowledge, and also better integrate, human cognition."
Researchers split decision-makers into three camps according to their style in making top-level business decisions:
- Skeptics: Executives who do not follow AI-based recommendations, and prefer to control decisions on their own.
- Interactors: Decision-makers who balance their own perception alongside AI input.
- Delegators: In a bid to reduce perceived individual risk, delegators shift decision-making authority to AI technology.
"When you integrate AI in these very senior executive decisions, you really have to acknowledge the fact that, as long as human judgment is involved, we are back to the same biases that we also have with AI," Meissner said.
AI guidance for business
The perception of an AI-enabled business attracts leaders. Applications promise efficacy and revenue by way of supplementing human labor, along with the ability to recoup investments.
There's been a link identified between digitally-savvy leadership and business performance.
Companies with digitally savvy leaders, those who are aware of emerging technologies and its impact, outperformed their peers on revenue growth and valuation by over 48%, another MIT research team found.
The appeal of AI has made its way to technology budgets, with the number of companies investing $500,000 to $5 million in AI rising 55% year over year, Appen's State of AI and Machine Learning report found.
But while technologists have previously identified strategies to combat bias in AI applications, what this research shows is that executive bias toward AI in decision making should also be addressed.
One recommendation is to enable team-based decisions, which allow the enterprise to combine multiple ways of thinking toward AI input on big business decisions.
"The biggest advantage of team-based decision-making, in our regard, is that teams consist of different people who have different decision-making styles," said Keding, a research associate at ESCP Business School in Berlin. "Because of that, we feel that this could also balance out the flaws in human judgment when dealing with specific AI-based advice."