202-450 test plan - LPIC-2 test 202, Part 2 of 2, version 4.5 Updated: 2023

ChatGPT is an artificial intelligence (AI) chatbot that uses a deep learning model to recognize patterns and relationships between words in its vast training data to generate human-like responses based on a prompt. But since there is no source of truth in its training data, the tool can generate responses that are factually incorrect.

"The use of large language models like ChatGPT is exploding and only going to increase," said lead author Rajesh Bhayana, M.D., FRCPC, an abdominal radiologist and technology lead at University Medical Imaging Toronto, Toronto General Hospital in Toronto, Canada. "Our research provides insight into ChatGPT's performance in a radiology context, highlighting the incredible potential of large language models, along with the current limitations that make it unreliable."

ChatGPT was recently named the fastest growing consumer application in history, and similar chatbots are being incorporated into popular search engines like Google and Bing that physicians and patients use to search for medical information, Dr. Bhayana noted.

To assess its performance on radiology board test questions and explore strengths and limitations, Dr. Bhayana and colleagues first tested ChatGPT based on GPT-3.5, currently the most commonly used version. The researchers used 150 multiple-choice questions designed to match the style, content and difficulty of the Canadian Royal College and American Board of Radiology exams.

The questions did not include images and were grouped by question type to gain insight into performance: lower-order (knowledge recall, basic understanding) and higher-order (apply, analyze, synthesize) thinking. The higher-order thinking questions were further subclassified by type (description of imaging findings, clinical management, calculation and classification, disease associations).

The performance of ChatGPT was evaluated overall and by question type and topic. Confidence of language in responses was also assessed.

The researchers found that ChatGPT based on GPT-3.5 answered 69% of questions correctly (104 of 150), near the passing grade of 70% used by the Royal College in Canada. The model performed relatively well on questions requiring lower-order thinking (84%, 51 of 61), but struggled with questions involving higher-order thinking (60%, 53 of 89). More specifically, it struggled with higher-order questions involving description of imaging findings (61%, 28 of 46), calculation and classification (25%, 2 of 8), and application of concepts (30%, 3 of 10). Its poor performance on higher-order thinking questions was not surprising given its lack of radiology-specific pretraining.

GPT-4 was released in March 2023 in limited form to paid users, specifically claiming to have improved advanced reasoning capabilities over GPT-3.5.

In a follow-up study, GPT-4 answered 81% (121 of 150) of the same questions correctly, outperforming GPT-3.5 and exceeding the passing threshold of 70%. GPT-4 performed much better than GPT-3.5 on higher-order thinking questions (81%), more specifically those involving description of imaging findings (85%) and application of concepts (90%).

The findings suggest that GPT-4's claimed improved advanced reasoning capabilities translate to enhanced performance in a radiology context. They also suggest improved contextual understanding of radiology-specific terminology, including imaging descriptions, which is critical to enable future downstream applications.

"Our study demonstrates an impressive improvement in performance of ChatGPT in radiology over a short time period, highlighting the growing potential of large language models in this context," Dr. Bhayana said.

GPT-4 showed no improvement on lower-order thinking questions (80% vs 84%) and answered 12 questions incorrectly that GPT-3.5 answered correctly, raising questions related to its reliability for information gathering.

"We were initially surprised by ChatGPT's accurate and confident answers to some challenging radiology questions, but then equally surprised by some very illogical and inaccurate assertions," Dr. Bhayana said. "Of course, given how these models work, the inaccurate responses should not be particularly surprising."

ChatGPT's dangerous tendency to produce inaccurate responses, termed hallucinations, is less frequent in GPT-4 but still limits usability in medical education and practice at present.

Both studies showed that ChatGPT used confident language consistently, even when incorrect. This is particularly dangerous if solely relied on for information, Dr. Bhayana notes, especially for novices who may not recognize confident incorrect responses as inaccurate.

"To me, this is its biggest limitation. At present, ChatGPT is best used to spark ideas, help start the medical writing process and in data summarization. If used for quick information recall, it always needs to be fact-checked," Dr. Bhayana said.

Phillip Martin / WGBH News/File

Under political pressure, a task force Tuesday changed its recommendations on a permanent admissions process for Boston's test schools in a way one of its leaders says will limit increases in their racial-ethnic diversity.

The task force will recommend 20 percent of the invitations to each of the city’s three test school be allocated to students by rank academic order city-wide, a shift from the policy the group had been hammering out over the last five months. Just a day earlier, the panel had reached a consensus that all invitations at each test school be allocated to students within tiers related to census tracts and socioeconomic status.

“The data is very clear about who and what the 20 percent represents,” said co-chair Tanisha Sullivan, who is also president of the Boston NAACP.

Sullivan said she didn’t want to make a concession, noting that the 20 percent of invitations benefit white students from more prosperous neighborhoods who have acess to greater resources to prepare to win admission. Black and Latino students have been underrepresented at Boston Latin School, Boston Latin Academy and John D. O'Bryant School of Mathematics and Science.

“I am a student of the civil rights movement, and so what I know to be true is that this is not new,” she told the task force. “We cannot be deterred. I want to be clear about this 20 percent conversation. It is political. As I look at the data it is clear about who this benefits … and there are those who are adamant it must exist.”

Neither Sullivan nor the other co-chair, Michael Contompasis, disclosed the exact source of the political pressure, but the Boston City Council will vote this week on the $1.3 billion schools budget and could reject it if a majority objects to the admissions plan.

Contompasis offered a way to lessen the impact of the last minute political demands on increasing diversity. He recommended giving students in public housing or foster care or who are homeless a better chance at admission by adding five points to their composite entrance score, a combination of their GPA and entrance test score.

A student's GPA will be weighted at 70 percent, and the test score at 30 percent. In the past, the two factors carried equal weight. Students attending a high poverty school where more than half of students are economically disadvantaged would receive 10 additional points to their score.

Contompasis, a former headmaster of Boston Latin School, noted the political sacrifice Sullivan made and acknowledged members have put in 60 hours of work since February to craft their recommendations.

But he also warned the members that if they did not agree to the change, “there may well be an impact on the district that none of us want to see."

The remaining 80 percent of test school invitations would be allocated to students within tiers under the latest task force recommendations, with the lowest socioeconomic tiers receiving offers first.

Boston’s test schools have not changed their admissions policies in two decades and temporary changes to the policies have already led to a fierce legal battle. Earlier this year, a group of white and Asian Boston parents sued the district for alleged discrimination saying the temporary changes in admissions to the test schools were racially motivated and discriminatory.

But U.S. District Court Judge William G. Young ruled in April that the plan was "race neutral" and allowed grades and zip codes to be used to assign students for the upcoming year. The school department said a traditional entrance test could not be administered this year because of the pandemic.

The 13-member task force had been charged with drafting a permanent plan that would expand diversity at the schools.

While several members said they would get behind the changes, four members said they disagreed or felt conflicted about the last-minute revisions.

Task force member Zena Lum, a parent of a student at Boston Latin Academy, expressed frustration. The task force’s charge was to create a more racially, geographically and economically diverse student body, and it had crafted a plan designed to accomplish those goals.

“We had a consensus, and it was taken away from us,” she said. “It has been taken away from us, not that it hasn’t been reached.”

Task force member Roseann Tung, a researcher, said the schools need more equity and justice, not 20 percent “set asides for the powerful.”

The panel ultimately agreed to present the recommendations to the school committee.

Sullivan said Superintendent Brenda Cassellius will present their recommendations to the school committee at its regularly scheduled meeting on Wednesday evening. The committee plans to vote on the recommendations in July. If approved, the changes would take effect in two years.

GBH News reporter Saraya Wintersmith contributed to this report.