You Have the Floor: IBM Watson AI Weighs In on Bloomberg’s “That’s Debatable”

Share this post:

WatsonIs it time to redistribute the world’s wealth? If you tune in to “That’s Debatable” on Bloomberg Television at 7 p.m. ET on Oct. 9, you’ll hear two teams of experts debate the topic in attempt to sway public opinion for or against the idea. You’ll also be witnessing, and hopefully participating in, a historical moment for artificial intelligence.

In an effort to bring even more global voices and opinions into the debate, IBM Research’s Project Debater team worked with the producers of “That’s Debatable” to integrate AI into the traditional debate format. With the help of a new natural language processing (NLP) feature called key point analysis, IBM Watson AI can synthesize thousands of submissions from the general public prior to each episode via The show moderator and host, Intelligence Squared U.S.’s John Donvan uses this information to incorporate public opinion into the debate.

Advancing the Language of Business

Key Point Analysis

“That’s Debatable” offers viewers an early look at key point analysis, a robust NLP technology already used in different contexts internally within IBM and available as part of Debater in our early access program, which gives clients the opportunity to use and evaluate different aspects of the technology. For the purpose of the Bloomberg TV show, we wanted to know how Debater would perform on a variety of topics across different episodes. This is a real exercise in the wild – we don’t have a say in the topics, and we don’t have any control over what viewers submit.

The original Debater system that debuted in 2018 relied on written material coming from thousands of sources to inform our AI’s point of view. The version of Debater technology featured on “That’s Debatable” focuses on arguments crowdsourced from people all over the world. The new version considers the quantitative nature of the output, as opposed to summarizing the information it analyzes into a narrative. The quantitative approach conveys the prevalence of each key point in the data analyzed—the number of arguments that support the point—whatever the words used to express the argument itself.

Key point analysis can examine thousands of documents—in this case, unedited submissions from viewers —and prepare a list of relevant points made in these documents by selecting, grading and filtering passages.Although the quality of the material analyzed varies, in theory it should be more on point than randomly searching the Web. You don’t have the issue of having to pinpoint arguments from billions of news articles and documents. But the information you do have to work with is noisier, and there’s the potential that some people submitting responses might try to troll the system by throwing in their own, unrelated arguments.

With key point analysis, we are getting closer to solutions that would be used by IBM customers interested in the quantitative nature of the expressed opinions. IBM is already integrating Debater technology into Watson. Since March, IBM clients using Watson Discovery, Watson Assistant and Watson Core Services have been able to take advantage of a number of features originally developed for Debater, including advanced sentiment analysis, summarization capabilities, advanced topic clustering and customizable classification of elements in business documents.



The Path to Prime Time

We introduced Project Debater to the world in June 2018 in the first ever live, public debates between AI and humans. The AI demonstrated it could quickly build a factual argument, consider a counter argument and deliver a rebuttal—all in everyday natural language. At the time, we saw Project Debater’s unveiling as the next step on a journey to deliver AI with NLP capabilities to businesses everywhere.

Debater’s technology has evolved with each public display of the technology. At the January 2019 CES conference, we demonstrated Project Debater Speech-by-Crowd decision-support capabilities. The following month, at our 2019 THINK conference in San Francisco, IBM Research hosted a third live debate for Project Debater, to test its growing skillset against champion debater Harish Natarajan. This high-profile live event introduced new levels of complexity. We had backups and systems controls in place to ensure reliability.

Last November at Cambridge Union, the world’s oldest debating society, Project Debater augmented two debate teams as they squared off by providing them arguments submitted by the public. For that event, Debater used our Speech by Crowd capabilities—the precursor to key point analysis—to demonstrate how AI could be used to enhance human reasoning.

“That’s Debatable” likewise follows the traditional Oxford-style format. In this case two teams, of two subject matter experts each, will debate over three rounds, and a live virtual audience will pick a winner via mobile app, to be announced at the end of the program.

With “That’s Debatable,” it is the first time we are giving a large number of people the ability to test Debater and contribute to the argument. We’re leaving the technology highly exposed so we can evaluate its performance under a new set of conditions. Our goal has always been to push the technology forward, making it a more versatile form of AI that would appeal to an ever-increasing audience of users.

Beyond this latest experiment, we will continue developing Debater with an eye toward the role it could have in people’s day-to-day lives. More than ever, people are expressing their views in public. Debater offers a way to provide a concise interpretation of those arguments. The technology’s relevance is more apparent as time passes and public discourse becomes more polarized.

The topic for episode two of “That’s Debatable,” to be broadcast November 6 at 7pm ET, is “A U.S.-China space race is good for humanity.” You can submit your argument here until Oct. 17.

Manager, IBM Debating Technologies

Noam Slonim

Distinguished Engineer, Project Debater, IBM Research-Haifa

More AI stories

State-of-the-Art Results in Conversational Telephony Speech Recognition with a Single-Headed Attention-Based Sequence-to-Sequence Model

Powerful neural networks have enabled the use of “end-to-end” speech recognition models that directly map a sequence of acoustic features to a sequence of words. It is generally believed that direct sequence-to-sequence speech recognition models are competitive with traditional hybrid models only when a large amount of training data is used. However, in our recent […]

Continue reading

IBM Research at INTERSPEECH 2020

The 21st INTERSPEECH Conference will take place as a fully virtual conference from October 25 to October 29. INTERSPEECH is the world’s largest conference devoted to speech processing and applications, and is the premiere conference of the International Speech Communication Association. The current focus of speech technology research at IBM Research AI is around Spoken […]

Continue reading

In Collaboration with the National Institutes of Health, IBM Research Dives Deep into Biomarkers of Schizophrenia

In collaboration with researchers from Harvard Medical School, Mt. Sinai School of Medicine, Stanford University and the Northern California Institute for Research and Education, IBM Research is undertaking a new research initiative funded by the National Institute of Health.

Continue reading