To explain or not? AI transparency depends on user expectations

Researchers created a simulated AI-driven dating website to investigate how user expectations influence the desire for transparency in AI systems. The study, which included Penn State researchers, revealed a direct correlation between how well the AI met or missed user expectations and the level of trust users placed in the AI system.

The findings highlight significant applications for industries such as healthcare and finance, where AI is increasingly used to streamline processes and enhance user experience. S. Shyam Sundar, co-author and director of the Penn State Center for Socially Responsible Artificial Intelligence, emphasized the impact on sensitive user interactions, saying, “AI can create all kinds of soul searching for people — especially in sensitive personal domains like online dating.” Users who received fewer matches than expected may feel inadequate, while those who received more might question their criteria.

The study involved 227 participants who used the fictitious dating site smartmatch.com and responded to varying conditions concerning the number of matches they were shown. Participants who received the expected five top picks reported trust in the system without seeking further explanations. However, when their expectations were exceeded or not met, they sought clarity, reinforcing the need for tailored transparency in AI.

The research indicates that as AI becomes more prevalent, companies must shift from standard disclaimers to user-centered explanations that actually enhance understanding and trust. By addressing user needs directly, industries can foster a more responsible approach to AI interactions, paving the way for improved user confidence and satisfaction.

The AI Journalism Labs at the Craig Newmark Graduate School of Journalism at CUNY, in collaboration with Microsoft and supported by the Nordic AI Journalism (NAIJ) network, is excited to introduce the 2026 AI Journalism Lab: Builders cohort. This select group of 24 professionals—including journalists, technologists, and product managers—will focus on developing AI-powered tools for newsrooms, addressing real-world challenges and enhancing operations.

Marie Gilot, Executive Director of J+ at the Newmark J-School, expressed pride in the cohort's expertise and anticipated innovative outcomes from this international collaboration. The program aims to empower participants to create AI solutions that improve internal workflows and engage audiences, prioritizing usability and ethical implementation.

Participants will work on hands-on projects, such as developing AI-driven applications for editorial processes, chatbots for audience interaction, and AI-generated content for storytelling. Melle Drenthe, an editorial innovation lead, emphasizes using AI to enhance the reporting process by meeting audience needs effectively. Notable applications include Annika Grosser’s AI solutions for editorial workflows at BBC News and Kalle Pirhonen’s automation of news gathering with AI tools at Ilta-Sanomat.

Additionally, Scott Klein from the Newspack team plans to help over 325 local publishers leverage AI for better community engagement and resource sharing. As the cohort convenes from January to May 2026, participants will pilot these innovations, ultimately aiming to strengthen journalistic integrity, enhance audience trust, and foster sustainable growth in news media.

Journalists play a critical role in reporting facts, a task made increasingly complex by the rise of artificial intelligence (AI), which facilitates content manipulation and the spread of misinformation. Hilke Schellmann, an investigative reporter and visiting professional at Princeton University's Center for Information Technology Policy (CITP), emphasizes the need for journalists to support society in distinguishing between fiction and reality.

During her fellowship at Princeton, Schellmann is exploring significant topics such as ageism in AI hiring practices. Collaborating with faculty, she investigates how AI tools like resume parsers might inadvertently bias age demographics. She advocates for educational initiatives to enhance digital literacy among students, asserting that understanding AI can foster critical thinking toward these technologies.

In journalism, Schellmann is focusing on AI accountability and how it transforms reporting. She is developing a toolkit for journalists to safely integrate AI into their investigative work, responding to demand for guidance in navigating AI applications. Schellmann also expresses concern about information integrity at a time when public trust in journalism is waning. She proposes the development of AI-enabled fact-checking tools to reinforce accuracy in reporting.

At CITP, Schellmann collaborates on workshops aimed at improving journalists' use of AI, addressing issues like ensuring the authenticity of sources in digital interactions. Her work reflects a commitment to leveraging technology in the public interest, emphasizing that while AI has significant applications in hiring and journalism, it also carries risks of bias and misinformation that need to be managed responsibly. As AI reshapes decision-making in critical areas, Schellmann calls for a more transparent and informed use of these powerful tools.

Following the recent U.S. operation in Venezuela, the use of artificial intelligence (AI) to generate content has surged dramatically. After Nicolás Maduro's capture, misleading images of him in custody quickly circulated on social media, many originating from AI tools designed to create realistic visuals. This phenomenon has created a significant amount of AI-generated media, highlighting both innovative applications and notable concerns.

On the platform X, a fabricated video showcasing crowds in Caracas gained a million views, while another AI-generated clip of purported celebrations amassed five million views, even catching the attention of Elon Musk. These videos, generated by platforms like OpenAI's Sora, illustrate a concrete application of AI in content creation, capable of transforming text prompts into compelling visual narratives. However, the authenticity of such content is increasingly questionable, raising ethical concerns.

Experts like Hany Farid from UC Berkeley observe that the unprecedented volume and sophistication of AI-generated media are altering the social media landscape. He estimates that the proportion of fake to real content has shifted significantly. This rapid proliferation stems from various sources: attention-seeking trolls, influencers, and even political figures. Although some creators do not intend to mislead, the presentation of AI-generated reenactments risks distorting public understanding of vital political events.

Darren Linvill from Clemson University expresses concerns that the credibility of these AI creations can shape real-world opinions and influence voters. Despite these challenges, Linvill posits that the demand for immediate news updates will spur even greater production of AI content, emphasizing the need for careful analysis in an era when "seeing is believing" may not hold true.