Meet Dolores, the chatbot helping people with chronic pain

August 10th, 2023

This piece, republished from The Conversation, explains how a CSIRO-developed chatbot is proving the adage 'a picture is worth a thousand words'.

Pictures of pain: how a visual chatbot can help people with chronic pain

David Ireland, CSIRO and Nicole Emma Andrews, The University of Queensland
The Conversation

Chronic or persistent pain is the main reason people seek medical care in Australia. Yet, most chronic pain is incurable and relies on education, exercise and other interventions for management, making it a complex, common and expensive health challenge.

The challenge is more pronounced in the treatment of children and people with diverse needs, such as those with language deficits and varying learning abilities.

Our team of researchers from CSIRO, RECOVER Injury Research Centre, the Tess Cramond Pain and Research Centre and the Queensland Interdisciplinary Paediatric Pain Service are exploring how therapeutic chatbots can help improve communication between people with chronic pain and health professionals.

To do this, we built a chatbot called Dolores, with a unique feature: a canvas for users to describe their pain with drawings, sketches and doodles. We found that in this case, a picture truly is worth a thousand words.

About Dolores

Dolores’s main function is to help users report data on their own pain and provide personalised education. This can be used to enhance communication between users and health professionals and improve pain management.

Dolores is designed to prompt the user for more information, ask follow-up questions, and generally lead the conversation.

Screenshots showing the Dolores app and drawing canvas.

The chatbot tailors its responses to the user. It will respond to children with short answers and a limited vocabulary, and responses will gradually increase in complexity with the user’s age.As the app often captures sensitive data, it was designed with privacy and ethical considerations in mind.

The technology behind Dolores is very different from the large language models behind chatbots such as ChatGPT. It works from a knowledge base and conversation flow built by occupational, speech and physio therapists, based on diverse experience with clients of different ages.

Dolores’s “brain” was built with help from hackathons and industry experts to ensure its responses are always focused on meeting the needs of users. Importantly, it will not give incorrect information or “hallucinate” responses.


A mind-map of some the conversation topics in Dolores.

Pain history

One thing Dolores is used for is to administer what is called a “pain history interview”.

In this process, Dolores asks questions about the onset, location and duration of the user’s pain. It also asks about the sensations and emotions associated with the pain and its impact on daily life, as well as the user’s understanding of pain and their perceptions of different kinds of treatment.

The user can respond to Dolores by talking, typing or drawing. Dolores then produces a transcript which the user can choose to share with their medical professional.

The transcript below was generated using speech recognition from a conversation with an individual living with chronic pain.

Dolores: Could you describe or draw what your pain physically feels like?

Client: Sometimes it’s different … it’s not all the same … then all the days it’s burning tingling

Dolores: Could you tell me or draw how your pain makes you feel? Physically and mentally.

Client: Shit makes me feel like shit

Dolores: And how long have you been in pain for?

Client: Since I was about 15 years old

Dolores: Can you tell me or draw how your pain began?

Client: I am not sure it just began

Dolores: And how does your pain affect your day-to-day life?

Client: It affects everything even down to driving, shopping, taking my kids to school. Even just making a cup of tea

Dolores: What do you believe is the cause of your pain?

Client: Autoimmune disease and follow on disorders from that

Dolores: What do you believe are the best treatments for managing your pain?

Client: Nothing really helps it’s just a matter of learning to deal with it and manage it

While this example was verbal and detailed, some users – particularly children and adolescents – preferred to doodle. For example, a volcano erupting symbolised a pain flare-up; lightning strikes symbolised sharp, sudden pain; and swirls and waves symbolised cyclic pain.

Sketches of clients given to Dolores during a chatting session on pain history.

When asking about the impact of pain on daily life, we received an abundance of stick figures and faces depicting sorrow and loss of identity.

Sketches of clients given to Dolores during a chatting session on the impact on daily living.

The future

In its present state, Dolores can interpret the colours used in a drawing, but not the drawing itself. Research shows red and black, for example, are commonly used when “painting pain”.

So when soliciting more information about a doodle, Dolores might say things like “That’s a lot of red. What is it?”

The next version of Dolores will have sketch recognition, which we believe will give her more insight and provide extended engagement with clients.

Dolores was received positively in cohorts of patients across different age groups. It will soon be used in the broader chronic pain intervention platform, Pain ROADMAP.

Dolores is a step forward in catering for clients and communities whose language may differ from the clinician and for children and individuals with language disorders, who often have higher rates of chronic pain and have difficulties communicating about their pain.

We believe a chatbot that identifies and conveys the symbols of pain to a clinician can be an effective tool for enhanced communication, leading to more effective pain management.The Conversation

David Ireland, Senior Research Scientist at the Australian E-Health Research Centre, CSIRO and Nicole Emma Andrews, Research Fellow, The University of Queensland

This article is republished from The Conversation under a Creative Commons license. Read the original article.