SPIKES Supercharged

LLM-powered website

We created a GPT-backed website that meant to help medical professionals better their empathy skills when breaking bad news.

Timeline
August 2023 - December 2023

User
Medical professionals/ trainees

Overview

Illustration of a doctor and a patient in a medical setting, with the doctor speaking to the patient who is sitting on a medical chair. The background has green tones, and there is a speech bubble with a check mark above the doctor.

SPIKES Supercharged is a GPT-4 powered training tool designed to help medical professionals practice delivering difficult news to patients with empathy. Built as part of my Human-Centered Computing graduate research, this project combines medical training frameworks with large language models to explore how AI can support empathetic doctor–patient communication.

Screenshot of a website titled 'SPIKES Supercharged' with the name letters SCIKES displayed horizontally and a chat input box at the bottom.
Screenshot of an online medical training platform titled 'SPIKES Supercharged' showing a step-by-step process with a highlighted step P, and a message about assessing a patient's knowledge and understanding during a difficult conversation.

Research and Backround

Research shows that while nearly all physicians recognize the importance of empathy when delivering bad news, fewer than half feel adequately trained.

Person typing on a laptop with a stethoscope on a wooden table.

The SPIKES protocol (Setting, Perception, Invitation, Knowledge, Empathy, Strategy) is widely used in medical training to guide this process.

X-ray image of a human hand making an OK sign.

We identified a gap: despite the rise of AI in healthcare, no tools yet existed that integrate LLMs to help professionals learn and practice empathy in high-stakes conversations.

Close-up of a patient's arm with a hospital wristband, holding to a hospital bed rail, in a medical setting.

Design Goals

Our goal was to create an interactive, user-driven system where medical students and professionals could practice SPIKES in a safe, repeatable way. We designed for:

  • Active learning through modular steps of the SPIKES protocol

  • Feedback, not answers, so users could iteratively improve

  • Flexibility to create custom patient scenarios based on real or hypothetical cases, able to practice areas as needed

  • Accessibility via a simple, web-based interface

A screenshot of a conversation between a medical professional and a patient about stage 2 breast cancer treatment options. The conversation includes instructions on incorporating SPIKES protocol and a supportive message from the healthcare provider.

Early design prototype

Process and Iteration

We developed a prototype chatbot that walks users through each step of SPIKES, prompting them to respond as if they were addressing a patient. The system then provides tailored feedback, highlighting strengths and suggesting improvements.

Early user feedback from clinicians and students highlighted the need for more meaningful, interactive feedback. Based on this, we refined the system to support repeated practice, scenario customization, and deeper guidance on difficult modules.

Diagram with six circles labeled S, P, I, K, E, S under the title 'SPIKES Supercharged.' Text explaining the second stage, P, stands for Perspective/Perception, and emphasizes asking patients questions about their understanding of their illness, their health condition, and screenings. Additional guidance encourages empathetic communication and understanding of patients' knowledge.

Early testing identified a need for more actionable feedback from tool

Red circular sign with white exclamation mark in the center.

User Experience (Scenario Walkthrough)

For example, a medical student named Sam practices telling a patient about a non-terminal breast cancer diagnosis. The tool guides Sam step by step—preparing the setting, gauging patient perception, explaining the diagnosis, and responding to emotions.

Screenshot of a medical history and patient information form for Mrs. Emily Rodriguez, including her age, gender, medical history, current health perception, recent test results, diagnosis, treatment plan, and family support details.

After Sam submits their response, the system evaluates it, gives constructive feedback, and allows them to retry until they are satisfied. This approach helps build SPIKES skills like muscle memory, while reducing the emotional load of real-world practice.

Screenshot of a webpage titled SPIKES Supercharged, with sections about patient communication, screening, and empathy, including highlighted text and a comments box.

Medical student Sam can practice any stage in SPIKES independently based on their desired need for practice

Screenshot of a medical training webpage titled SPIKES Supercharged, showing a step-by-step guide for communicating with patients, with a focus on assessing patient knowledge and understanding.

User Testing and Reflection

We deployed an online survey to investigate user perceptions. The tool measured high in helpfulness, usefulness, and moderately well in interaction success.

"I think this would be great for medical students or students in other advanced practice programs (NPs, PAs, CNMs)."

— Anonymous User, User testing

This project taught us how to translate medical communication frameworks into interactive learning tools, while balancing technical implementation with user-centered design.

Future iterations could incorporate richer feedback loops, adaptive roleplay, and integration into medical training curricula. Ultimately, this project raised exciting questions about how AI can support—not replace—empathy in healthcare.

See the rest of my portfolio
See the rest of my portfolio