Connect with us

Hi, what are you looking for?

Tech & Science

Op-Ed: Project Debater – IBM’s AI can argue effectively with humans

IBM's Watson™ is a question-answering computer system capable of answering questions posed in natural language. - Tim Sandle
IBM's Watson™ is a question-answering computer system capable of answering questions posed in natural language. - Tim Sandle

Project Debater, a descendant of IBM’s Watson, IS a major technical achievement. The degree of cognition and coherence required for AI to formulate a working argument is truly vast. Logic and objectivity are required to manage information in a formal debate. The video is from 2019, but is a good example of how Project Debater works.
This AI needs to persuade and deliver a comprehensible argument to humans. Try finding that combination in a herd of press releases. It ain’t easy. In the video, Project Debater argues in favor of subsidizing preschools The gigantic range of this subject and its correlative issues, like society and societal needs, is pretty daunting.
Project Debater does a good job of building an argument. (Please note the range of information Project Debater uses in the debate, another strong factor in the development and credibility of truly sentient AI.) The human debater tries to defuse the AI’s arguments. It’s worth watching, even if you have strong views on this subject, and I do. I’m very pro-subsidy.
The interaction of the debate is also interesting, counter-arguing points.
That requires presence of mind, and more to the point, presence of effective arguments.
A new business culture, a new range of metrics, or both?
IBM have made a bigger cultural point than they may realise with Project Debater. Much human activity is data-driven, and Project Debater is in effect delivering a form of data, if not in the usual way.
The culture of metrics is unambivalent. It’s the interpretation of metrics that’s often disingenuous. It’s easy to argue with people, and easier to argue with numbers, in theory. In practice, metrics and clearly expressed arguments DO disturb the outcome. Not because they’re necessarily persuasive, but because they’re documented and unavoidable.
The most inept decision is harder to make with clear information proving the opposite of the arguments in favor of that decision. Even spin, the standard evasion technique, doesn’t do well.
Imagine this very basic scenario:
• AI delivers a clear argument for a specific business decision.
• The human argument is for an alternative decision, against the AI argument.
• The final decision for either case depends on meeting agreed business objectives.
• The argument against the AI view is clearly weak and equally clearly serves a personal interest rather than a business interest.
• Numbers for the AI argument stack up much better than the private interest selective figures.
In effect, the AI has exposed an unworkable option in the alternative which is being sold to the business as a future direction or outlay. Management may or may not be overly adept at specific areas of the issues, but the AI has clarified the situation.
The likely result is that management can now safely refuse the human-sponsored alternative due to its weaknesses. Note the word “safely”. The safety lies specifically in a clear argument in favor of the decision. That’s the epitome of playing safe in management culture.
All information is supported by facts and much more relevantly in the context of stated or unstated business objectives. No prizes for guessing which argument is likely to appeal to management, or why.
Business culture, in fact social culture, needs valid argument. Arbitrary or autocratic decisions are by definition made by excluding argument. One look at this world shows you where that quality of decision gets.
(The video points out that many people may never have really heard counter arguments, let alone clear arguments. That’s one of the truly sick things about this world as a whole; excluding the legitimate interests of so many people.)
Project Debater could also be used as a professional training tool to help professionals understand counter-arguments. This type of training simply cannot be turned into yet another rote learning exercise.
Preset views and ideologies can’t work, because the information provided won’t, and can’t, run on that basis. Facts don’t have to fit theories, let alone ideologies, which rarely fit everyone and sometimes don’t work for anyone.
I’ll be interested to see how Project Debater evolves. This is a very strong argument for AI – Clean information, clear argument, and quality control on decisions. Try arguing with that.

Avatar photo
Written By

Editor-at-Large based in Sydney, Australia.

You may also like:


United Airlines reported higher quarterly profits Wednesday but announced it would trim US capacity later this year.


Wearing goggles and a wet suit, the 65-year-old city leader swam breaststroke before immersing her face and beginning a front crawl.


A landmark European Union law known as the Digital Markets Act (DMA) entered into force in March.


Scientists are remaking the things people want and need using net zero technologies.