NEWS
NORMAN, Okla. – With artificial intelligence upending how people seek, interpret and act on information, efforts are underway to design AI systems that are equitable, efficient and inclusive.
A University of Oklahoma data scientist has created a free research tool to facilitate this process.
Called ECHO – Evaluation of Chat, Human Behavior, and Outcomes – the open source, low-code platform enables scholars to design and run behavioral experiments involving conversational AI, Web search and human-AI interaction.
“Our platform saves researchers time and money in front-end programming,” said Jiqun Liu, an associate professor specializing in data science at the OU School of Library and Information Studies. Liu created ECHO at his Human-Computer Interaction and Recommendation (HCIR) lab with OU graduate student Nischal Dinesh and Ran Yu, a senior researcher at GESIS – Leibniz Institute for the Social Sciences, in Germany.
“When a researcher wants to build a study that integrates an AI chatbot, they can install our platform and easily add components,” Liu said. “For most social science studies, ECHO will be easy to implement.”
Appetite for analyses of human-AI interaction is growing, said Liu. Not long ago, keyword search engines were the dominant way people found information online. Today, conversational AI-powered “answer engines” are increasingly delivering data, changing how people acquire and use information.
But researchers are still studying how these shifts influence learning, trust, and decision making. Studies incorporating human-AI interaction can help shed light on these developments.
To use ECHO, researchers must first install the software onto their own computer. The code is available for free on GitHub, and a step-by-step video tutorial guides users through the setup process. Installation takes about 20 minutes.
Once installed, researchers can configure experimental workflows through an administrator dashboard, while participants complete consent forms, surveys, and information-seeking tasks through a web-based interface. The platform records behavior logs and participant responses, which can be exported as structured datasets for analysis.
Liu said he uses ECHO for several collaborative research projects currently underway. One examines how much information people retain when they use an AI chatbot for research and learning on debated topics. That project is being conducted in partnership with GESIS.
Another study, with colleagues at OU Health Sciences, is testing whether AI tools can improve online information access and interactive advice related to intimate partner violence.
By making the ECHO platform free and available to anyone, Liu said he hopes more researchers will implement AI-human interaction designs into their work.
“ECHO enables researchers to configure complete experimental workflows through a flexible administrative interface rather than code,” the researchers wrote in a paper introducing the system. “ECHO lowers technical barriers, promotes methodological consistency and enables cumulative, interdisciplinary research on human centered evaluation of modern information access systems.”
ECHO’s origins trace to a previous research project funded by the National Science Foundation, which looked at people's biases and cognitive limits in their interactions with non-AI powered Web search engines.
In his lab, Liu studies information retrieval, human-AI interaction, and behavioral science to understand how people engage with Web search engines, conversational systems, and large language models (LLMs) during information seeking and decision-making processes.
Partial funding for the transition to ECHO came from a $15,000 OU Faculty Investment Program award.
Like any piece of software, ECHO is constantly being updated. To gather improvement feedback, Liu said he’s encouraging researchers at OU and elsewhere to use the platform to design their experiments – and share their experience doing so.
Ultimately, Liu said his goal in creating ECHO is to encourage “healthy friction” in AI system design and evaluation. Past research suggests that because AI tools and chatbots are often overly agreeable, people are more trusting of inaccurate, biased, or misleading information. The first step in critically assessing what a chatbot produces is a deeper understanding of how boundedly rational humans and AI engage.
“There are problems with human-AI interaction,” said Liu. “As an information scientist, I can't solve all these problems. But what I can do is develop open-source systems, like ECHO, that others can use to boost their own study of these interactions.”
About the University of Oklahoma
Founded in 1890, the University of Oklahoma is a public research university located in Norman, Oklahoma. As the state’s flagship university, OU serves the educational, cultural, economic and health care needs of the state, region and nation. For more information about the university, visit www.ou.edu.
A University of Oklahoma data scientist has created a free research tool to facilitate this process. Called ECHO – Evaluation of Chat, Human Behavior, and Outcomes – the open source, low-code platform enables scholars to design and run behavioral experiments involving conversational AI, Web search and human-AI interaction.
Three students at the University of Oklahoma have been awarded nationally competitive scholarships, including two Fulbright recipients and one Gilman awardee.
The Shyam Dev Patwardhan Department of Philosophy at the University of Oklahoma is outperforming national benchmarks for doctoral student success, according to newly released 2026 data from the Academic Placement Data and Analysis project.
