Reasoning with Machines? Okay.

Updated: Apr 23

Artificial intelligence is now capable of forming arguments and engaging in debates against humans. This could be the first step in helping us make better decisions.

AI helping humans to make better decisions

Sifting through and making sense of the immense volume of information in today’s content-driven world is a daunting endeavour, especially with an estimated 1.145 trillion MB of data produced daily.


More than ever, we are relying on search engines such as Google to cut through the clutter, but as they get better at generating relevant results, this process of personalising and targeting content has inadvertently limited our exposure to issues and opinions that differ from our own. The result? We grapple with synthesising increasingly polarised viewpoints.


Consequently, developing the ability to think critically has become crucial, not just in classifying and assessing the accuracy of information, but also in evaluating information to create convincing logical arguments.


Such critical thinking skills were once thought to be beyond the purview of artificial intelligence (AI) because of it’s limitations in understanding context, nuance and emotion used in human reasoning. The structure and flow of these arguments are also still poorly understood by researchers, making the development of computer systems that can identify and create arguments using natural language extremely challenging.


A possible breakthrough


Despite the challenges, a number of laboratories worldwide are working on advancing argument technology, taking advantage of the technical developments in AI, the increase in the amount of data for training, and the strong commercial demand for this type of technology.



Harish Natarajan, who holds the world record for most debate competition victories, took on IBM Project Debater, the first AI system that can debate humans on complex topics, at a live debate at IBM Think 2019.
Credit: IBM Research on Flickr

One of these attempts is the IBM Project Debater, an AI-powered autonomous system that can engage in complex debate with humans. The system was trained with large data sets covering around 100 debate topics. Then, using a process called argument mining—analysing a vast database to identify statements that provide supporting evidence for a claim—Project Debater was able to present a case supporting a chosen issue. It performed admirably against world class opponents, winning two thirds of the debated topics.


The success of Project Debater signals a turning point for the technology, but there are still a few more challenges to overcome before it can be applied in the real world. Problems such as navigating argument webs of “cross-references, analogy, exemplification and generalisation”, becomes even harder when you remove the limits of a structured debate. The real acid test of any AI argument technology will be its ability to tackle societal-scale debates.


The future of decision making


As we learn more about the intricacies of how arguments are formed, AI can provide useful insight into what makes or breaks arguments. With the ability to analyse arguments, filter out irrelevant data, account for biases, and categorise large amounts of information into “for” and “against'' evidence, it can help improve a quality of debate by highlighting weak points that humans tend to overlook.


As more quality data is obtained for training algorithms, and as scientists develop new methods in teaching computers about how arguments work, AI’s ability to create and even critique arguments will further evolve to something that more closely mirrors human reasoning.


With these new developments in the field of AI, and as we learn more about the intricacies of how arguments are formed, it can help us to make better decisions around critical issues, such as foreign policy and climate change, that can have significant impact on the world.


By taking conflicting views into account and considering the justifications behind them, AI can lead us to take a broader view of issues, instead of being confined within our own viewpoints. This can result in better-informed and more well-rounded arguments, which ultimately lead to better decisions. This can be especially valuable when examining issues that are complex and can be emotionally-charged. National security, equality, climate change – these are all conversations that are significant as they are emotionally-charged.


Being optimists, we work towards a world where AI can be the solution to the havoc that has been wrought by tribalism, making arguments that instead of dividing, contribute to more productive, thoughtful, and reasonable discussions.



Main Source:

Reed, Chris. “Argument Technology for Debating with Humans.” Nature News, https://www.nature.com/articles/d41586-021-00539-5

Other Related Sources:

Hornigold, Thomas. “How Do You Win An Argument? IBM’s New AI Has a Formula.” Singularity Hub, https://singularityhub.com/2018/07/05/how-do-you-win-an-argument-ibms-new-ai-has-a-formula/

Reed, Chris. “How AI Can Make Us Better at Arguing.” The Conversation, https://theconversation.com/how-ai-can-make-us-better-at-arguing-85938

Heaven, Will Douglas, “IBM’s Debating AI Just Got a Lot Closer to Being a Useful Tool”, Technology Review, https://www.technologyreview.com/2020/01/21/276156/ibms-debating-ai-just-got-a-lot-closer-to-being-a-useful-tool/

Reed, Chris. “The computers being trained to beat you in an argument.” BBC News, https://www.bbc.com/news/technology-41010848