Two professors in the Pratt School of Engineering pioneered a personal AI research engine whose machine learning models are tailored for use in academia.
The program, Inquisite, was developed primarily by Jon Reifschneider, executive director of the AI for Product Innovation master’s program, with assistance from Pramod Singh, executive in residence in the engineering graduate and professional programs. It officially launched to the public Oct. 17.
Inquisite works similarly to other language models and search engines, such as ChatGPT or Google, but specializes in conducting research on complex topics. Its main appeal is a promise to give users responses generated from trustworthy sources.
Unlike other models, Inquisite provides citations for all sources it uses in its responses, as well as relevance and reliability rankings for each source.
“If you go to Google and type in your question, you’re going to get a bunch of stuff, and half of it’s going to be people's blogs about their opinion … or maybe misinformation or maybe conflicting information, and you’re not sure which to rely on,” Reifschneider said. “Inquisite does that for you and prioritizes the use of trusted sources.”
According to Reifschneider, Inquisite identifies credible sources by assigning trust scores to academic websites. These scores depend on factors such as number of citations and whether the source was published.
Reifenschneider and Singh first began the project to help researchers find important information in their fields and optimize their time.
“[A researcher’s] results would be as good as the first part, which is the search part,” Singh said. “Your search has to be good and bring in the relevant information.”
After speaking to potential users, the professors realized that Inquisite’s improved search return feature could be used by a larger audience than just academics.
“For the vast majority of people, they either don't have the time to read [research papers], or they don't have the capacity, because they're not a trained expert in that field.” Reifenschneider said. “What we're ultimately trying to do is make that type of trusted information easily accessible.”
To launch and market their product, members of the Office for Translation and Commercialization provided Reifenschneider and Singh with support to set up the technology.
Following its initial launch, Inquisite is taking feedback from users and using it to grow. So far, such input has yielded a feature that allows users to follow up on questions they ask the model.
Singh and Reifenschneider are already planning for the tool to access and judge non-academic sources and are incorporating more factors into trust scores to enhance their reliability.
“We're also right now working on integrating other types of signals, like the tonality of the article,” Reifenschneider said. “Is it written in sort of a factual, objective manner, or is it written in a sensationalized manner? Does the source itself cite other sources as references in it, or is it just free text that somebody sat down and wrote based on their own opinions?”
Ultimately, the founders hope to use Inquisite to help make information more accessible.
“We do believe this can help democratize and revolutionize research,” Singh said.
Get The Chronicle straight to your inbox
Signup for our weekly newsletter. Cancel at any time.
Winston Qian is a Pratt sophomore and health/science editor for the news department.