Patient-centered emotional care, how agents alleviated the psychological burden on cancer patients

Mental health care is often treated as a secondary priority in most clinical settings. Immediate physical treatment takes precedence, and mental health assessments are typically conducted only at specific points in time—often limited to one-time surveys. In many cases, the system relies on patients to recognize their own condition and actively seek help.

However, most patients do not speak up first. By the time others are able to recognize a problem, the patient’s condition has often already worsened.

With the use of agents, patients can be supported in monitoring and managing their own mental state. Instead of ending with a single assessment, continuous interaction enables ongoing evaluation and care, allowing patients to be connected with medical professionals before their condition deteriorates.

Genesis Lab is developing a distress management agent for cancer patients. Distress in cancer patients differs from general mental health issues. Cancer patients experience simultaneous physical, psychological, and social changes brought on by treatments such as chemotherapy and radiation therapy. This complexity has led to the emergence of psycho-oncology as a specialized field. However, in reality, it is difficult for all patients to receive intensive care.

In Korea, approximately 1.74 million people have been diagnosed with cancer, and an estimated 35–45% of them experience clinically significant levels of distress. Many patients remain in need of care but are unable to receive it.

Replicating the Clinical Approach of Specialists

Pebble is not an app that introduces a new model of care. Instead, it translates the clinical approach of psycho-oncology specialists into an agent.

The agent replicates the actual consultation flow and decision-making process used by specialists, implementing them as an algorithm. The scenario is structured around a holistic assessment framework:

📌 [이미지: 시나리오 흐름도] 인사 → 정신과적 증상·신체적 증상·생활습관 평가 → 종합 평가 → 결과 설명·지지 면담 → 치료 연계·마음챙김 기반 인지치료(MBCT)

Importantly, the symptom evaluation process is not a static questionnaire. The agent dynamically generates follow-up questions based on the patient’s responses. When physical symptoms are reported, it explores that area in greater depth; when emotional expressions arise, it continues along that context.

Pebble’s counseling scenarios have been validated and refined using 229 real consultation cases conducted by three psycho-oncology specialists affiliated with Seoul National University Hospital, Eulji University Hospital, and Inha University Hospital. This close collaboration with clinicians enabled us to design an agent that is practical and applicable in real clinical settings.

Enabling Patients to Monitor and Manage Their Own Condition

Pebble is built on Agentria. Conversations with the AI counselor—from greeting to treatment referral—are modularized into distinct components.

Creating a safe space for expression
Patients engage in conversations with an AI counselor, allowing them to express thoughts and emotions that may be difficult to share with medical professionals. Through context-aware dialogue, the agent encourages patients to articulate both emotional and physical experiences. Based on the patient’s responses, the agent continues to ask follow-up questions, assessing psychiatric symptoms, physical conditions, and lifestyle factors together.

Below is an excerpt from an actual AI counseling session:

AI Counselor: How do you feel about the possibility of cancer recurrence or progression? If you do have such concerns, could you share how strong they are?
Patient: Of course I’m worried about recurrence. I’m currently undergoing treatment and hope everything goes well, but I’ve heard many cases where cancer returns even after treatment, so it’s hard not to worry.
AI Counselor: That’s completely understandable. Many patients share similar concerns. Along with that, have you ever felt worried about becoming a burden to your family or loved ones?
Patient: It’s hard not to feel that way sometimes. I do feel like I might be a burden. But I try to remind myself that we’re family, and it’s okay.
AI Counselor: I see. It sounds like you’ve been carrying concerns about being a burden to your family. At the same time, your effort to reassure yourself reflects a deep sense of care and love for them.From fear of recurrence to guilt about family, the agent follows what the patient says and continues the context. It’s not a fixed script; it understands the conversation, empathizes, and proceeds with counseling.

From fears of recurrence to concerns about family, the agent understands the patient’s responses and continues the conversation within context. Rather than following a fixed script, it engages in empathetic, context-aware dialogue.

Making one’s condition visible
The content expressed through conversation is structured according to a diagnostic framework. Patients can view their previously scattered symptoms in an organized format, helping them better understand their own condition. These records are also documented for clinicians to review in future consultations.

Tracking changes over time
The process does not end with a single session. Records are connected over time, allowing both patients and clinicians to track changes and trends. Patients can reflect on whether they are improving and identify periods when they experienced greater difficulty.

Connecting to care at the right time
When a high-risk level is detected, the agent recommends seeking professional care. In the future, attending physicians in oncology departments will be able to directly review these results and request collaborative care with psychiatry.

📌 [이미지: 환자용 결과 확인 화면]

Real Patient Experience

A prototype was tested with 31 cancer patients, including individuals with both high and low levels of depression. The results showed that all participants were able to use the system smoothly, regardless of their condition.

This indicates that the tool is accessible to both emotionally vulnerable patients and those in relatively stable states. Importantly, patients perceived the service not as a simple psychological test, but as a meaningful tool for understanding their condition.

In particular, the Mindfulness-Based Cognitive Therapy (MBCT) content received the highest satisfaction score, averaging 4.52 out of 5. This suggests that the experience of actively managing one’s own mental state—beyond passive assessment—can function as a valuable form of care.

Cancer patients often carry a significant burden throughout their treatment journey. Physical changes, emotional fluctuations, and the restructuring of daily life are difficult to fully express during limited clinical encounters.

Pebble aims to create a system where patients can continuously monitor their own condition and connect to appropriate care when needed. It is not intended to replace specialists, but to extend care into the gaps where specialists cannot always reach.

Pebble demonstrates how agents can be designed for real-world application. By deeply understanding the context and constraints of a specific domain, and by building systems that integrate with human decision-making, agents can evolve beyond simple tools into solutions that address meaningful problems.