What does AI mean for Equity in Debate?

Ever since the launch of ChatGPT in 2022, Artificial Intelligence has been integrated into everything from Kahoot to the browsers we use everyday. With everyone taking sides, debate has been slow in choosing whether to allow AI or not. Nonetheless, if you walk into a motion preparation session nowadays, you'll likely see at least one computer with a Large Language Model website open. Debate is no exception to the impacts of AI, and the crucial point to answer now is whether these tools will have a positive or negative impact in shaping our discussions.

One of the first times I interacted with these technologies was last year. AI was all over the news, and so, my coach decided that we would debate against it to see how much of a threat it raises. He gave us a motion and we made cases following our usual routines. When it was time to speak, he recorded what we said and then fed it to the AI for it to generate refutations against our points. It responded exceptionally well, giving refutations that only experienced debaters could come up with, but then it started hallucinating information and evidence. At first I was intrigued, asking myself how I could apply it elsewhere, and then I became alarmed. Would AI give other teams unfair advantages in competitions? What would be the impacts of its use?

Looking back, many of these questions seem exaggerated, but the question of what AI will do to the debating community still remains. Artificial Intelligence has already revolutionized how we learn, with many learning platforms integrating this new technology into their tools. For example, Duolingo uses AI for personalized feedback and Khan Academy has automated math tutors that you can chat with. In the past, learning how to debate involved reading manuals, watching videos, going to practices, and writing cases. Now, with the advent of AI, debaters can skip many of these steps, instead opting for applications powered by these technologies that promise to improve their speaking skills quickly. 

One example of these applications is Speech Coach AI. This application works by prompting you to record a speech and then feeding it to a LLM for it to give you feedback. It offers you a selection of five different formats, but lacks support for World Schools. In the testing that I conducted of this application, it gave me solid feedback about my points and was helpful, but it did not have any feedback on style, since it only analyzes a transcript of what you say. Its website claims that it is not a replacement to traditional coaching, but their advertising seems to contradict this message by presenting itself as a more accessible option.

Speech Coach AI is not the only application that has been created with the mission of helping you improve your speaking, in fact, a whole market exists with this objective. Orai, Astound and Vocal Image are just some other examples of applications that have the same promise. These applications are directed more towards professionals who wish to improve their speaking skills, and Speech Coach AI is directed towards debaters. The mission of these applications may seem appealing to many, as they claim to make speech and debate accessible to all. But limitations to this arise on two main points. The first one is on profit. All of the applications I mentioned previously have at least one model of monetization that requires the user to spend their money. The second limitation is on availability. As of right now, only a handful of these applications are available on app stores outside of the United States, limiting their benefits to a small part of the world. 

With this, it’s clear that AI can assist you in improving your speaking skills, but, another major part of how debaters use these tools to assist them is for generating arguments and ideas. The majority of people use ChatGPT for this, but it has some limitations. Because it is a Large Language Model, it is focused on providing answers to the widest range of topics possible. This and the fact that it was trained on information from the internet mean that it has limited knowledge on debating concepts. For example, it has limited information on how to impact arguments or on how they should be structured. This raises issues for debaters who employ these tools because it means that they get arguments which are either badly written, have broad scopes, or are based on false information. This not only raises issues in accuracy and accountability by creating debates focused on false claims, but also leads to lower-quality discourse.

To try and solve this problem, I wanted to see what an AI that was custom-made for debating would look like in comparison to ChatGPT. For this, I used a website called Poe to create an AI chatbot that focuses on writing cases for the World Schools format. Poe is a website where users can create chatbots by setting their prompts, which are the instructions that the AI follows. Following the website’s recommendations for writing prompts, I created one that told the bot how it should structure its arguments and what key concepts it should use. And like that, Worldie was born. To use the bot, you need to provide it with a motion and a side of the house, and it will create a case following the SEAL format. The chatbot is, of course, also plagued by many of the issues faced by other AI-powered applications. It hallucinates information frequently and can give you arguments that are not important enough or that go against your side, but it is an interesting exploration of what custom-made AI solutions for debate could look like.

A real-life example of AI-powered tools focused on the writing process for debate is Contention AI. This application aids debaters in cutting cards with its machine learning technology. Cutting cards is a practice commonly done by members of the speech and debate community in which sources are stripped down to the key facts that are useful for the debate. You give the application a topic and it will generate 10 cards based on sources it finds on the internet. This tool is developed by a company called Koios and like many of the speech applications mentioned previously, it has a subscription model for access to the platform.

The emergence of AI in debate has benefits: It makes learning more accessible and it reduces the strain put on them while researching or coming up with arguments. Like all things, these benefits come at a cost, with that being lower-quality debates based on false information in some instances, and less creativity in general. The problem that AI raises is in fairness. Not everyone has access to high-speed internet, and not everyone knows how or when to use these tools. This issue is prevalent in the United States, with 42 million citizens not having access to broadband according to Forbes, but it is exacerbated in developing countries like Mexico, where only 14% has this kind of access according to the OECD. Even if people in developing countries have access to high-speed internet needed to use these tools, they would still need to pay the subscription fees needed to have unlimited access to responses and feedback. 

Other issues that are raised with the use of AI in case writing are that it may defeat the whole purpose of debating. Debate is supposed to be an activity in which you learn how to think critically and argue. The use of tools like Worldie or Contention AI could potentially eliminate this part of debating because they complete the debater’s job for them. On the other hand, debaters who use these tools could learn how to interact with them in face of a future in which it is constantly becoming more prevalent. The upsides and downsides of the use of these tools are things that each debater will have to weigh when deciding whether or not to use them.

The condoning of the use of AI in competitive debate may increase inequality and may lead to unfair competitions, or it could give millions more the opportunity to develop their speaking skills and ease debater’s workloads. We must carefully manage AI's integration into debate as it becomes more and more entwined with it to make sure it improves our conversations without sacrificing accessibility, fairness, or quality.