SIU School of Medicine invests $2M in AI avatars to teach future doctors how to deliver bad news

By WAND · November 6, 2026.

The Springfield Memorial Foundation and the Southern Illinois University (SIU) School of Medicine are investing $2 million to develop AI-based medical training programs. The work is being carried out at the Ed Curtis Laboratory for Innovation in Medical Education in Springfield, Illinois, according to local station WAND.

The technology turns real people into digital avatars. Medical professionals train them to reproduce typical patient interactions and questions, and students converse with them via camera and microphone.

One of the core platforms is a difficult-conversations simulator. According to Pinckney Benedict, the program's director and the lab's chief architect, it focuses less on diagnosis and more on the relationship with the patient and on learning to deliver bad news with empathy. Scenarios include referring a patient to hospice care, informing them of a new allergy or delivering a cancer diagnosis.

According to Benedict, the cases come directly from physicians. The lab has also begun working on correctional medicine, focusing on medical issues at Illinois Department of Corrections facilities. In addition, the platform lets doctors create their own cases: they simply write a description, for example “a person with lung cancer,” and the system generates the case. The lab aims to share the program with other medical schools across the country.

Our take: the most significant detail is not the technology but the problem it tackles. Delivering a grave prognosis is one of the most demanding skills in medicine and, as Benedict says, it is barely trained: it is learned “cold,” with real families. A simulator allows students to make mistakes, repeat and receive feedback at no human cost. It is a clear case of AI augmenting human capability rather than replacing it.

Caution is also warranted. The source offers no evidence of outcomes: there is no data showing that students trained with avatars communicate better than those trained with actors or traditional simulations. Nor do we know how the quality of an avatar's empathy is assessed. These are the questions that will determine whether this is a pedagogical advance or a flashy demo. In general, simulation with standardized patients (actors) already exists in medical schools; what's new is the near-zero marginal cost of each repetition and the ability to generate cases on demand.

The broader implication fits our long-term thesis. If AI helps train better-prepared and more accessible physicians, and does so at scale, the benefits extend beyond the major universities. In the short term, the risks are the usual ones: dependence on proprietary platforms, bias in generated cases and the temptation to cut supervised human training to save costs. The right bar is measuring outcomes, not demos.