Renée van der Nat, researcher at the ‘Quality Journalism in Digital Transition’ professorship at Utrecht University of Applied Sciences, is involved in the Disrupted Connection project, which was launched this year. It must show what role chatbots and smart speakers can potentially play in strengthening – or regaining – trust between journalism and the public. All through so-called conversational artificial intelligence.
Van der Nat does not have a smart speaker at home, she admits. Not a disaster: the current generation of devices is not even technically capable of carrying out an ongoing conversation, AI or not. Also, to be honest, they’re not very smart. By 2026, when the research ends, speakers will probably be able to do more.
However, Verbroken Verbinding emphatically does not focus on these electronics, but should mainly provide practical recommendations for journalistic topics that are suitable for an AI conversation. And thus help to achieve greater trust in journalism.
Three prototypes
Over the next two years, Van der Nat, together with colleagues Yael de Haan, Winnifred Wijnker and Sophie Duvekot, will investigate which topics are suitable for a journalistic conversation. Verbroken Verbinding must also soon produce at least three working prototypes, according to Van der Nat.
The project also involves news organizations in these practical tests and in determining suitable journalistic topics for conversational artificial intelligence. The national NOS, the regional channel Omroep West and local channel Omroep Venlo participate in Verbroken Verbinding.
Other partners are the Dutch Local Public Broadcasters Foundation (NLPO), the Fontys Journalism and Design professorship and the University of Groningen. Joris Heijkant, who is also involved in AI development for Villamedia and various other journalistic titles, focuses on the development of the technology behind AI-driven conversations.
In addition to prototypes, the research should also provide practical basic principles that editors can use when developing their own conversational artificial intelligence applications.
Conversational artificial intelligence or a conversation with artificial intelligence is literally that: an ongoing conversation via a chatbot or smart speaker. Current smart speakers and chatbots actually only respond to specific questions (the weather, the time, the news), but conversational artificial intelligence, just like a human speaker, remembers what has been said and responds in context with appropriate follow-up questions.
Like all AI, it relies on enormous amounts of data, (language/speech) models and machine learning. The latter means that the AI learns from previous experiences (or mistakes) without being specifically programmed for this.
This should ultimately lead to a more natural conversation. Research must be carried out by the Lectorate over the next two years Quality Journalism in Digital Transition demonstrate how feasible such a conversation is and which journalistic topics are suitable for it.
The popular rejection bots in various customer service departments or the speakers of Google Home or Amazon Alexa are primitive examples of conversational AI but do not really deserve the label ‘intelligent’ for the time being.
Tags: Research started conversational bridge journalism distrustful public
