Volunteers Practiced Laughing at Their Own Troubles With a Chatbot Coach, and Gains Held 3 Months – Medical Daily

Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
How to Live Better, Longer
Most people wait for something funny to happen before they laugh. A research team at Imperial College London is testing whether people can learn to generate humor on their own, even about their own difficulties, with coaching from a chatbot.
In a pilot study published March 13, 2025, in Frontiers in Digital Health, 27 adults who completed an eight-week program reported large improvements in well-being that remained three months later. The same team is now running a randomized trial to see whether the effect holds up against a comparison group.
Abbas Edalat, a professor in Imperial’s Department of Computing, developed the program, called the Self-Initiated Humour Protocol, with colleagues from computing and the Faculty of Medicine.
It focuses on non-hostile humor. Psychologists often sort humor into four styles: self-enhancing, affiliative, self-defeating, and aggressive. The first two are considered non-hostile, and the protocol aims to build them using theories of why people laugh, including incongruity, superiority, and play. Unlike most laughter programs, it does not require a group and is designed to be practiced alone.
Its digital helper was a web-based, rule-based chatbot rather than a large language model, a choice the team made to avoid toxic or made-up responses. It was designed to sound empathetic and humorous and to detect a user’s emotional state. The team started with 122 base responses and expanded them to 2,507 written by volunteers and crowdsourcing. A separate AI model, which identified humorous text with 97% accuracy in testing, helped pick which responses to use.
Participants rated the chatbot’s tone positively, describing it as humorous, empathetic, and polite, and found it easy to use. They were less impressed by the conversations themselves, citing a rigid flow and repetitive phrasing.
The program also used virtual reality. Each participant received a personalized avatar of their younger self, built from a favorite childhood photo.
For the first two weeks, participants practiced a technique in which they built a compassionate bond with their childhood self through the avatar. For the next six weeks, they worked through 10 humor exercises, with at least 20 minutes of daily practice recorded in a diary and weekly Zoom sessions led by Edalat.
The exercises ranged from feigning laughter and laughing at oneself to spotting incongruity in the world and laughing at difficulties. One example in the paper encourages people to laugh as a sign of victory after finishing a dull chore like washing up or cleaning.
Thirty-two adults enrolled, and five dropped out. To join, volunteers had to score below 15 on standard screening questionnaires for depression and anxiety, which excluded people with more severe symptoms. Seventeen of the completers had mild, or subclinical, symptoms.
Compared with their starting scores, participants showed large improvements in overall well-being, self-compassion, the ability to reframe stressful situations and the use of self-enhancing humor, which involves finding the funny side of one’s own troubles. Psychological capital, a measure that includes hope and resilience, improved moderately.
Those gains remained at the three-month follow-up. The subclinical group showed similar or larger improvements. The other three humor styles, affiliative, self-defeating, and aggressive, did not change significantly.
The authors acknowledge the biggest limitation: the study was not randomized and had no control group, so its findings are only preliminary. Without a comparison group, it is impossible to tell how much of the change came from the program itself. The sample was also small, the outcomes were self-reported, and the weekly sessions were presented by Edalat, who developed the protocol. Participants attended an average of about 5.8 of 9 sessions and completed about 40 of 56 diary entries.
The authors say the results show potential for medium-term benefits that should be validated in randomized trials.
That trial is already in progress, with one notable change: after two weeks with the childhood avatar, participants practice with a chatbot coach powered by Meta’s Llama 3 large language model. They also watch a weekly prerecorded presentation of new exercises and fill out questionnaires before the program, at eight weeks, and at three months.
According to its ISRCTN registry entry, it is a single-center, single-blinded waitlist randomized controlled trial in adults with subclinical symptoms, funded by the Empowered Human Foundation in the U.K. Recruitment closed on Sept. 30, 2026, with 46 participants enrolled, and the study is expected to finish at the end of November 2026. As of the registry’s latest update on Oct. 5, 2026, no results had been posted.
The switch to generative AI raises the stakes. In November 2025, the American Psychological Association warned that even generative AI tools built with sound psychological science lack enough evidence to show they are effective or safe in mental health care, and that chatbots and wellness apps should not replace care from a qualified professional. The Imperial trial is the kind of controlled test that could start to supply that evidence for humor coaching.
For now, humor training remains a promising wellness idea, not a treatment. People dealing with persistent low mood or anxiety should talk with a qualified clinician. In the U.S., anyone in crisis can call or text the 988 Suicide and Crisis Lifeline.
What is the Self-Initiated Humour Protocol?
An eight-week program from Imperial College London that teaches people to create non-hostile humor themselves rather than waiting for jokes.
Did the pilot use ChatGPT or a similar AI?
No. The pilot used a rule-based chatbot with thousands of prewritten responses, plus a virtual reality avatar of the user’s childhood self. The follow-up trial uses a Llama 3-based chatbot.
What did the pilot find?
Among 27 completers, well-being, self-compassion, emotional reframing, and self-enhancing humor improved, and the gains held for three months.
Why are the results preliminary?
There was no control group or randomization, the sample was small, and the outcomes were self-reported.
Is a larger study happening?
Yes. A waitlist randomized trial enrolled 46 people and is expected to finish in November 2026. No results have been posted yet.
Can this replace therapy?
No. It was tested in people without severe symptoms, and anyone struggling should seek help from a qualified clinician.
Published by Medicaldaily.com
© 2026 Medical Daily Inc. All rights reserved
Medical Daily is for informational purposes and should not be considered medical advice, diagnosis or treatment recommendation. Read more.

source

Scroll to Top