“We need to get used to manipulation as a society” | Facts and Contexts Matter | bpb.de

Jeanette Hofmann, Freie Universität Berlin, and Masato Kajimoto, University of Hong Kong, discussed the spread of misinformation and disinformation in Europe and Asia in the opening session of the conference.

How does generative AI change information environments and media literacy interventions?

The AI models we have in place reflect the regulations and ethics of the places where the systems are built. “At the moment, our AI models are mostly in line with what US companies think we should value”, said Antonio Krueger, CEO and scientific director of the German Research Center for Artificial Intelligence, “and Chinese models are in line with how Chinese regulators think they should work.”

Krueger reflected more generally on the power of generative AI and ways to include, for example, watermarks in photos and videos to distinguish human from artificial content. The other two panelists spoke of local applications using generative AI. In South Korea, the popular chatbot Iruda, which had been strongly criticized after disseminating hate speech, was tamed by introducing a generative AI model. “The first version of Iruda was trained by 100 million chats retrieved from the “Science of Love Service” (a platform on relationship advice). Her hate speech came from chats humans exchanged”, recounted Sungook Hong, Seoul National University. He researched how Iruda was transformed and explained how the tech start-up Scatter Lab, which programmed Iruda, established ethical principles in co-creation with the ICT Policy Institute and Iruda’s users.

Use artificial networks to stimulate human neuro networks – this was the message of Isabel Hou’s input, who is the Secretary General of the non-profit organization Taiwan AI Academy, which aims to educate industrial stakeholders on AI technology trends. In her talk, she introduced the fact-checking bot Cofacts, which runs on open-source technology and whose database consists of knowledge crowdsourced by active users. Hou stressed that generative AI alone is no silver bullet: “Empower readers to enhance their digital literacy so that they can make their own decisions.” And Krueger added: “We need to get used to manipulation as a society – and learn new tools.” Tools, which are informed by a diverse range of stakeholders.

The second panel focused on the challenges of generative AI. In addition to describing existing trends, the experts presented local applications - such as the Korean chatbot Iruda and the Taiwanese chatbot project Cofacts.

Talking politics: State regulations and ideologies

On August 25, 2023, the EU’s Digital Services Act (DSA) officially came into power. Tech giants like Google, Meta, and Amazon are now being held to stricter standards of accountability for content posted on their platforms and need to include feedback options for their customers to be able to report problematic content. Mauritius Dorn from the think tank Institute of Strategic Dialogue explained how the DSA works and recounted some of the discussions that led to the Act aiming to foster safer online environments. He talked about problematic areas: “We need to be aware of what elements we as European Union are focusing on in our cyber diplomacy.” Jeanette Hofmann, Freie Universität Berlin, who moderated the session, added: “The problem is that it is up to the nation-states to define illegal content and thus what platforms need to take down.” This is especially crucial in states where courts are not independent. The DSA could end up as a powerful tool for EU countries that restrict the right to freedom of speech.

While Europe wants to regulate large platforms with the DSA, Japan relies more on the work of non-governmental organizations, to combat misinformation and false information. In the third panel Matthias Dorn, ISD, and Daisuke Furuta, JFC, explained why and discussed the advantages and disadvantages of (non-)regulation.

Daisuke Furuta, editor-in-chief of Japan FactCheck Center claims that this may be one of the reasons the Japanese government is against platform regulations. “In Japan, the study group on platform services of the Ministry of Internal Affairs and Communications recommended that the private sector should combat misinformation, rather than it being regulated by law.” Organizations with a “neutral governance structure” should be strengthened. “Of course, private fact-checking organizations cannot deal with big tech platforms alone”, noted Furuta and mentioned the need for some kind of regulation. “However, this is not a priority of our current government.” Overall, it seems that Japan is less affected by misinformation and disinformation – like for example its neighbor Taiwan, where Furuta’s fact-check organization has close ties with like-minded organizations: “Most fact-checking tools are from Western countries, but we lack for example Chinese language tools. Taiwan is at the forefront of developing these.” According to him, more knowledge of “non-Western” platforms like the Chinese microblogging service Weibo or the Japanese instant messaging app LINE is needed to combat misinformation or harmful speech.

A look at four countries: How to deal with misinformation and hate speech?

Harnessing global perspectives for collective solutions

The last panel specifically looked at the connection between misinformation and hate speech and started off with a question about the definition of hate speech. What actually is hate speech? “In Germany, the constitutional prominence of human dignity shapes our approach to hate speech”, said Mauritius Dorn, ISD. “In South Korea the definition focuses on discrimination. Hate speech means that people are excluded from mainstream society”, added Sookeung Jung, CCDM. Isabel Hou, Taiwan AI Academy, senses a strong focus on disinformation that could be linked to hate speech, however, the courts in Taiwan define what hate speech is. Masato Kajimoto, Hong Kong University, raised the problem of politized courts in some countries in Asia, and Dorn added that platforms have plenty of scope to design their products and policies, which is why more transparency of content moderation and recommender systems is urgently needed. The DSA tries to contribute to this endeavor.

In the final panel, the experts mainly discussed the link between misinformation and hate and possible counter interventions.

The panelists agreed that a holistic approach is necessary to tackle hate speech. “Legal regulation is not enough. There needs to be behavioral regulation, that directly prohibits hate speech from being expressed, for example in classrooms. Also, digital environments need to be modified so that it becomes harder to engage in hate speech”, said Jung. Kajimoto outlined different ways that classrooms could address this problem. Some address only what’s being said. Others say: “Let’s learn about the history, let’s learn about the psychology of hate, let’s open wounds.” The panelists noted that media literacy education should address people of all age groups. Approaches for tackling hate speech require good coordination and the engagement of the whole society. For Dorn it’s also about “innovating democracies” via new tools and spaces in the public interest: “We are in a state of multiple crises, so we need spaces where people can get fact-based information about the impact of issues on their lives and ask their questions without hate and agitation.” 

All in all, the conference showed that Asian and European countries have more in common when it comes to dealing with misinformation than perhaps initially assumed.

Interner Link: Conference program and speakers in PDF format