From fact-checking to argument-checking as Award Nominated researchers of University of Amsterdam join MMGA with human-AI framework.
The spreading of disinformation and fake news is dividing the world increasingly and we are seeing more impactful societal and psychological consequences each day. Fact-checking is a key action nowadays, but it doesn’t cover all aspects of debunking misinformation, disinformation, even fake news: verifying the way facts are used within arguments should be part of it. But there are scientific solutions coming up! Two of the initiators of KRINO, Dr. Federica Russo and Dr. Jean Wagemans, envision KRINO as a source of transparent, explainable AI that helps users to analyse and evaluate arguments. The software can be used to unmask fake news but also verify the reliability of the information for example needed to build complex systems such as an airplane.
The KRINO team members of the University of Amsterdam and the ISOC Make Media Great Again Working Group (MMGA) decided to collaborate in order to strengthen each other’s mission and form a stronger community to reverse the trend of disinformation and fake news.
The arguments KRINO analyzes can occur in various contexts – from political manifestations and speeches, to doctors’ recommendations, to social media posts. In contrast to many AI-based tools, KRINO will not be a fully-automated engine, since that would defy the purpose of serving as human aid rather than dictating the outcomes based on algorithms purely. The AI-based tool is in development and will be largely based on the learnings that the KRINO team will gather by collaborating with the MMGA Working Group community.
Beyond fact-checking
“One should also take into account the relationship between facts and the specific points of view that, for instance, politicians propose, or the policies they want to convince people of.“
Verifying the way facts are used within arguments is the gap KRINO is trying to bridge.
Jean: In an interview about the KRINO project published in the Dutch National newspaper Volkskrant last year we emphasized the importance of fact-checking for improving the quality of information. Fact-checking has always been useful and important as an internal procedure for journalists. In recent years it has evolved into a published item, with the trustworthiness of public actors and newspapers depending on it.
However, in my view fact-checking wouldn’t be enough to make the media great again and inform citizens about scientific results or validated information. In an article I wrote together with my colleague José Plug, we propose to extend this activity to rhetoric-checking, considering more than just facts and including an assessment of arguments and rhetorical strategies.
In our view, one should also take into account the relationship between facts and the specific points of view that, for instance, politicians propose, or the policies they want to convince people of.
It was good to meet Ruben (editorial; Ruben Brave is the founder of MMGA): we all want to work on something in order to change the status quo, but not necessarily fact-checking and AI based. We both want to focus on the involvement of human annotators.
Why now?
“Studying rhetoric is not a panacea but it will help to know even a little bit about what types of arguments there are and how to evaluate them. It will empower people to defend themselves.“
The urgency of working on these matters now is that we have to act before it gets more out of hand, seeing the scale of false information spreading right now and the vast societal consequences it’s already having. What happens if we don’t work on this?
Federica: The optimistic scenario, as I view it, is that we will continue to produce poor quality content. The pessimistic scenario instead is that this will get worse and worse. I’m thinking a lot about it. It feels like David against Goliath with MMGA plus KRINO representing David: the Internet is far too big. And then I thought: ‘No, this is a chance!’. If we don’t do it now we miss the opportunity to try and counteract a tendency. Besides: inaction is also action.
Jean: It’s a difficult issue, the current misinformation and fake news creation and quick spread and its negative societal effects, but I hope that more and more people are at least aware of the fact that we have a problem. That’s the first step.
What might help in creating this awareness is to know that the problem we are facing is not new, on the contrary. People have always been misled by means of rhetorical tricks. There is this ancient Greek graffiti, scratched in a wall with a nail, saying: ‘Whoever does not study rhetoric will become a victim of it’. If you know how this part of our communication works, then chances of you being fooled by people are considerably smaller. Studying rhetoric is not a panacea but it will help to know even a little bit about what types of arguments there are and how to evaluate them. It will empower people to defend themselves.
New ways of teaching
The team behind KRINO also has an aim to find new ways to teach people about argumentation and rhetoric and empower them with critical thinking skills. Eventually that will provide people with a shield for all aspects of life, but also very much for their onlife (online life).
Jean: I’m excited about the collaboration with MMGA. First of all we’re talking about a different audience. Usually I am only teaching students, and I’m looking forward to sharing my expertise with other groups. And second, I agree with Federica that basically it’s about enhancing people’s rhetorical literacy. Because of this collaboration we will be introduced to and develop together new ways of empowering people. There is an opportunity here to radically change the way in which you motivate people to look at these problems.
We are used to teaching students from textbooks. It’s the format we use. The exciting idea is: are there other means to teach people to work with tools that help them evaluate arguments in texts? There’s no doubt that there are different ways in which people take up information and learn new things, let’s experiment further with that.
Mutually beneficial
All parties involved in this collaboration benefit from it: it’s a three-way street. For the KRINO team the collaboration with the ISOC MMGA Working Group is about learning from the process. The ISOC MMGA Working Group community can benefit from the webinars or online courses that the KRINO team will develop and present, on how to apply argument-checking to various sources. The KRINO team meanwhile learns from the community what type of difficulties they face and which sources they see as particularly challenging. The MMGA team adds to it by developing the community further and gathering more annotators to the platform, thereby stimulating the KRINO-team to dynamically scale up their classes and learning-modules. At all levels it is clear however that the pool of annotators is central to the success of both the KRINO and MMGA team.
Besides the societal mission and training, there’s a practical side to the collaboration with the KRINO team. It will greatly help them develop KRINO as an AI-based tool, by feeding it many annotations and making the software learn from it. Argumentation theory is so contextual that the tool needs to also learn to understand various contexts and what types of arguments one can encounter there, in order to consequently aid in validating them.
Jean: This tool, KRINO, is an envisioned AI. We are not there yet. We have just started in January with part of the linguistic implementations. The next step will be to enable the engine to detect, identify, and assess arguments on the basis of the Periodic Table of Arguments.
That’s why we’re interested in the annotations already done by MMGA, and interested in designing training for annotators, so they can learn how to add an extra layer in their annotations, that of argument-checking besides fact-checking.
Consequently, if we continue our research on how annotations work and the pool of annotators grows, we can finetune the findings and adjust the procedures. Only then we will start thinking of implementing a pseudo-algorithm into the machine.
Human-centered development approach
The focus of the KRINO team lies on developing a human-first AI tool. The ethical implications are continuously put in the forefront, during the entire development process. Part of that philosophy is to create an explainable tool, not a black box as many AI tools that are currently being produced.
Federica: Our view for KRINO, since the beginning, is that it should be an aid to humans. We never want it to be fully automated. We are working with engineers and a linguist to build an AI to analyze texts and the arguments in it and there is the prospect of automatizing argument evaluation. But to me the question is whether it’s able to reflect what we want. How do we want a machine to help a human user to evaluate arguments? How can we use this technology responsibly?
It’s a slow, interdisciplinary process. But the interesting fact is we’re thinking ahead about what we would like this thing to do for us. Ethical considerations don’t come at the end as a cherry on the cake. They’re built into the design.
What’s next?
“The urgency of working on these matters now is that we have to act before it gets more out of hand, seeing the scale of false information spreading right now and the vast societal consequences it’s already having. What happens if we don’t work on this?“
KRINO and MMGA will now focus on developing a webinar and training course for annotators of the MMGA Working Group.
Stay tuned to hear more about the latest developments around the upcoming MMGA- KRINO training!