IBM computers have beaten humans at Jeopardy! and chess. But the company’s artificial intelligence system was unable to beat a champion debater – at least according to a human audience.

The AI system, called Project Debater, faced off against 2012 European Debate Champion Harish Natarajan in California on Monday. Both were given the debate topic with just 15 minutes to prepare.

Project Debater, a tall black box with a female voice, scanned a database of hundreds of millions of newspaper and magazine articles and wrote a four-minute argument based on what she found.

She was asked to argue in favour of the statement “We should subsidize preschools.” Natarajan argued against.

Project Debater’s argument included a torrent of research and statistics purporting to show that preschool graduates have better outcomes when it comes to things like education, health and crime.

Natarajan, meanwhile, argued that subsidizing preschool is a “politically-motivated giveaway to members of the middle class” and that there are better ways to spend the money.

Project Debater said in her rebuttal. “For starters, I sometimes listen to opponents and wonder, what do they want? Would they prefer poor people on their doorsteps begging for money?”

“The state budget is a big one,” she went on. “The idea that there are more important things to spend on are irrelevant.”

Natarajan pounced on that part of her argument. “She notes that maybe the state has the budget to do all the good things,” he said. “I would love to live in that world but I don’t think that is the world that we live in.”

Before the two squared off, 79 per cent of audience members agreed with the statement “we should subsidize preschools,” while 13 per cent disagreed and eight per cent were undecided. Afterwards, the proportion in agreeance dropped to 62 per cent, while 30 per cent disagreed and eight per cent were undecided. On that result, Natarajan was declared the winner.

However, Project Debater was given the edge when the audience was asked to vote on the question: “Which one of the two debaters better enriched your knowledge?”

IBM computer scientist Dan Lahav told CTV News Channel just ahead of the debate that he believes Project Debater shows how computers can complement human decision making.

“Project Debater is quite good at scanning and finding relevant arguments, matching them with evidence in order to ground them and make them much stronger, being able to construct a speech which is coherent and relevant based on that,” he said.

“Obviously humans have a lot advantages such as understanding the relevance, seeing the audience ... humour etc.,” he added.

IBM has said that “the goal is to build a system that helps people make evidence-based decisions when the answers aren’t black-and-white.”