October 20, 2026
Ann Liu is a PhD candidate in the Institute of Communications Research, which houses the world’s oldest doctoral program in communications and media. Liu’s research interests focus on emerging technologies like GenAI and the social impact of data and quantification.
Describe your areas of research interest.
I study emerging technologies, data and quantification, and persuasion in strategic communication. My research follows two main lines: emerging technologies, focusing on generative AI (GenAI), especially AI-generated summaries, and it also examines the social impact of data and quantification, focusing on numbers in persuasion.
My recent projects examine how AI-generated summaries transform media exposure and message creation, as well as how consumers perceive and respond to AI-generated summaries across various contexts, such as generative search engines, online review platforms, and social media.
In addition to AI-generated summaries, my research on GenAI also examines how individuals seek information with GenAI, how GenAI shapes creative processes and outputs, and persuasion outcomes.
I explore how companies can use numbers responsibly and ethically in message design to reduce the risk of misleading audiences, as well as how individual traits (e.g., numeracy) help audiences identify misleading uses of numbers in persuasive messages.
Why does your research matter?
My first research line focuses on GenAI, which is transforming the information we produce and consume in everyday life. Think about searching for brands and products using Google. In addition to the brand’s official website, e-commerce platforms’ product pages, and other individual sources, we now have AI-generated summaries of search results, which aggregate multiple pieces of information from diverse sources into a single piece of information. Also, think about checking out how other consumers evaluate a product or service. AI-generated summaries are now deployed on online review platforms to synthesize multiple consumers’ reviews.
Indeed, AI-generated summaries involve information from multiple sources, thereby may present more than one perspective to the audience. Additionally, AI-generated summaries streamline the volume of information, thereby reducing the required cognitive resources for consumers to get to know about brands and products. However, what sources are selected and what information gets represented in the summary are algorithmic decisions made in that “black box”. One of my works (Liu & Zhang, 2026) used computational methods to empirically audit AI-generated summaries in representing sentiments of public discussion on social media, finding that the algorithm tends to select positive posts from all social media posts as “reference posts,” and generate a summary that is even more positive than the reference posts. In other words, AI-generated summaries may be biased in representing the original information through selecting certain sources and reflecting certain aspects of information.
Building on this work, my ongoing projects focus on the transformations that AI-generated summaries bring to media exposure (which sources get selected?) and message creation (which information is reflected in the summary?). As a media technology and strategic communication researcher, I believe this technology brings up critical questions for both academia and the industry.
Be cautious when relying on AI-generated summaries, as GenAI may not be as smart as you in strategically selecting appropriate sources and critically evaluating information.
Anna Liu
ICR Doctoral StudentWhile a PhD student here, what work/experiences are you proud of?
I’m proud of my research accomplishments, as well as the teaching experiences that have allowed me to build meaningful connections with my students.
During my time in ICR, I have published four first-authored papers so far, with another co-first-authored paper in press. I also have multiple first-authored papers and collaborative works currently under review. My work has appeared in journals such as New Media & Society, Computational Communication Research, Convergence, and Advertising & Society Quarterly. Beyond my research output, I’m also proud of the meaningful collaborative relationships I have built with faculty members and fellow graduate students in ICR.
In addition to research, I have had valuable opportunities to serve as the instructor of record for Advertising Research Methods. I’ve also served as a teaching assistant and led discussion sections for numerous advertising courses. These experiences have allowed me to build meaningful connections with students, support them in their coursework, and help them consider their future career opportunities.
How does your research help consumers?
The two focuses of my research are rooted in people’s information-seeking in everyday life.
One of these focuses has been relevant for a long time, as the use of quantitative data in message design is not new. In everyday life, consumers frequently encounter advertising claims such as “9 out of 10 dentists recommend our brand.” However, when consumers treat numbers simply as heuristic cues or scientific evidence, they may be misled. For example, where did the data behind the claim come from? How was the survey question designed? What if the question allowed multiple responses and another competing brand was recommended by 10 dentists? How large was the sample, and who were the dentists surveyed? Without this contextual information, consumers who simply accept the claim that “9 out of 10 dentists recommend our brand” may miss a competing brand that could better meet their needs.
The second focus of my research examines an emerging phenomenon driven by the massive amount of information available in the digital age. We now have more information than ever before, yet our cognitive resources remain limited. As a result, it is increasingly convenient to use generative AI to summarize news articles, compare brand features, or synthesize information from multiple sources. However, this convenience raises an important question: To what extent should we trust GenAI? Can we assume that AI is capable of selecting the most appropriate information sources and determining which aspects of that information should be represented in the answers it provides? My research seeks to understand how consumers navigate these questions and how they make decisions when AI becomes an intermediary in their everyday information-seeking. In other words, be cautious when relying on AI-generated summaries, as GenAI may not be as smart as you in strategically selecting appropriate sources and critically evaluating information.
What led you to be interested in your research area?
The critical thinking I developed through my industry experience, combined with the inspiring phenomenon happening in the constantly changing media landscape. My professional experience taught me to question how data and quantification can potentially mislead people. This also provides a perspective that motivates me to view algorithms and GenAI, which are built on probabilistic models and trained on data. They are often perceived as objective or scientific technologies. Yet, as human-created artifacts, these technologies can reflect biases depending on how they are designed, trained, and used, and by whom.
In contemporary society, our attention is limited, while the volume of information continues to grow. As we increasingly rely on numbers to understand information and on GenAI to provide condensed and streamlined information, we need to remain cautious and thoughtful. We should keep our human intelligence involved by critically evaluating, verifying, and questioning the data and technologies we encounter.
Share on social