The Epilepsy Algorithm

CONTEXT

 

In 2022, as part of the Unknownness Lab, we began laboratory work around the topic of mathematical algorithms, in particular Artificial Intelligence, for the prediction of epileptic seizures. The goal of this scientific area, in the future, will be the integration of algorithms in devices in order to predict seizures, communicate them to the patient and, ideally, disarm them through neuromodulation mechanisms.

 

This Laboratory work is carried out with the collaboration of researchers from the Centre for Informatics and Systems of the University of Coimbra, where research in this field is being developed.

 

One of the researchers, who is part of the project team, explains the topic thus: “Imagine you are a doctor looking after patients whose medication does not completely suppress the epileptic seizures. A researcher presents you with a device that, by reading the brain’s electrical activity, can predict seizures. In practice, a patient is driving, hears the device’s alarm, stops the car, has a seizure and the harm is minimised. But apart from the fact that the device can fail, there is another problem: its mathematical operations are so complex that even the researcher cannot humanly explain them. Would you, as doctors, give the device to your patients?”

 

 

ARTIFICIAL INTELLIGENCE

 

This tool generates the ability of machines performing complex tasks with characteristics that we associate with intelligence (learning, reasoning, inference, pattern recognition). Artificial Intelligence is present in multiple aspects of our daily life: social networks, online sales mechanisms, risk mitigation systems (bank fraud detection, hiring individuals, insurance contracting), computer games, military strategy, translation systems, aeronautics.

 

In its most common form, as in crisis prediction, this intelligence is acquired by supervised learning algorithms: we give an algorithm input-output pairs, and it builds a model to, with new inputs, guess its output. In this scientific area, researchers provide the algorithms with data of brain electrical activity from both pre-crisis and steady states, with the aim of building a model that can distinguish between them.

 

Learning from examples is a very simple but extremely powerful idea. Do we learn from examples? If we give all the existing text in the world, in input-output pairs, to an Artificial Intelligence algorithm, will it be able to learn everything? It’s amazing how ChatGPT works: based on the question we ask, it gives us the most likely word to continue the sentence. It then adds that word to the original question, forming a new input to the model, which then gives us the next most likely word as an answer. It repeats this process until the end of the answer. In conclusion, it is hard to tell if ChatGPT “thinks” the whole sentence before saying it, maybe it is so, but its reasoning is word by word. Isn’t it crazy?

 

 

MEDICINE

 

Medicine is one of the areas where artificial intelligence has been applied for some decades, namely in diagnostic support tools. Returning to the previous question: is it acceptable to make a high risk, medical decision through the use of mechanisms whose “reasoning” we cannot fully understand? What is the acceptable margin of error for the use of these devices? What is the limit for a wrong decision to be a lesser evil compared to the potential advantages? What does it mean for a person with epilepsy to have their seizures predicted by a machine connected to their brain?

 

The experience and perception of a patient are elements that we intend to integrate in the creation process. On a methodological level, we organised discussion sessions bringing together artists, scientists and clinicians working in the field. We also conducted interviews with doctors, patients and carers, collecting their experiences and perspectives on this neurological disease and the possibility of using devices like the ones described.

 

For the organisation of the interviews and scientific advice, we were supported by the Refractory Epilepsy Reference Centre of the CHUC. We defined a script of questions, submitted to the ethics board of the CHUC, which were the basis for semi-structured interviews. The collected testimonies were then used as reference material for the dramaturgical creation.

 

 

THE EPILEPSY ALGORITHM (Show)

 

There are several ethical issues that arise when working with these kinds of themes. However, at the same time, they are great material for dramaturgical exploration, since they touch on contemporary and future concerns, but also on universal and timeless human dilemmas.

 

“The Epilepsy Algorithm“, produced in close partnership with Teatro da Cerca de São Bernardo, in Coimbra, is the result of a collaboration between Marionet and the Department of Informatics Engineering of the University of Coimbra, being one of the ideas that was awarded funding by Feedzai‘s First Foundation Project, a company created by faculty members of the Department of Informatics Engineering. The essays and the writing of the text had contributions from researchers at the Centre for Informatics and Systems of the University of Coimbra, people with epilepsy and health professionals.

 

 

THE EPILEPSY ALGORITHM (Post-show Conversation)

 

After the last presentation session of the show “The Epilepsy Algorithm“, on July 2nd, 2023, Teatro da Cerca de São Bernardo hosted a conversation informally addressing the various themes that inspired this play and that found artistic expression in it. In addition to the participation open to the public, the panel of guests included several professionals that work on scientific research, medicine and theatre, as well as a person suffering from epilepsy.

 

As during the show’s creation process, public participation was highly valued, so we invited all the participants to raise questions or enrich the debate with personal stories and experiences. This was a moment of personal and scientific sharing of great relevance, taking into account one of the main goals of the show which happened immediately before, the increase of general knowledge about epileptic disease and other associated topics, through theatre and artistic creation.

 

 

CLOUDS – MARIONET DIGITAL (Digital Artistic Object)

 

As part of this project, we created “Clouds“, a video that embodies the second expression of Marionet Digital, which aims to be a space for artistic expression, created in formats designed for digital communication. This digital artistic object deals with the theme of epilepsy, in which we imagine what might happen in the head of a person with this illness.

 

Epilepsy is a disease of the central nervous system that causes abnormal electrical discharges in neurons. It is known that 1% of the world’s population suffers from it, regardless of gender or age. In addition to the seriousness of the disease itself, given the risk of accidents, trauma, prolonged seizures and sudden death caused by convulsions, it is also directly related to psychological (anxiety, depression, etc.) and social (social isolation, unemployment, etc.) conditions. However, with advances in medicine, it is possible to control the disease and reduce its side effects, promoting the health and well-being of people suffering from epilepsy.