Genetic artificial intelligence changes the way humans read, speak, speak, think, empathize and act within and across languages, cultures. In healthcare, communication gaps between patients and practitioners can worsen patient outcomes and prevent improvements in practice and care. Language/AI incubators are made possible through funding from the MIT Human Insight Collaborative (Mithic) and provide a potential response to these challenges.
The project envisions a humanities-based research community to promote interdisciplinary collaboration across MIT and to deepen understanding of the impact of generation AI on interlinguistic and intercultural communication. The project focuses on healthcare and communication, and aims to bridge the entire socioeconomic, cultural and linguistic stratum.
The incubator is co-led by Leoceri, a physician and research director, research director and senior research scientist at the Institute of Medical Engineering Sciences (IMES), and professor of German and Second Language Research Practices and director of MIT’s Global Language Program.
“The foundation of healthcare delivery is knowledge of health and illness,” says Seri. “We see poor results despite the large investment due to broken knowledge systems.”
Chance collaboration
Urlaub and Celi met at the Michik launch event. Conversations during event receptions have been revealed to have a common interest in investigating improvements in medical communication and practice with AI.
“We are trying to incorporate data science into our healthcare delivery,” says Seri. “We have been recruiting social scientists (at IMES) to help us move forward in our work, as the science we create is not neutral.”
Language is a neutral mediator in health care delivery, and the team believes it can be a benefit or barrier to effective treatment. “Later, after we met, I joined one of his working groups, whose focus is the ratio phors of pain: the language used to explain it and its measurements,” continues Urlaub. “One of the questions we considered was how effective communication occurs between doctors and patients.”
Technology affects casual communication, and its impact depends on both users and creators. As AI and large-scale language models (LLMs) gain power and excellence, their use is expanding to include areas such as healthcare and wellness.
Rodrigo Gameiro, a physician and researcher at MIT’s Institute of Computational Physiology, is another program participant. He points out that he works at a laboratory center responsible for developing and implementing AI. A nuanced approach is needed to design systems that effectively utilize AI, especially when considering communication-related challenges across linguistic and cultural divisions that may arise in healthcare.
“When building AI systems that interact with human language, we don’t just teach words, we teach them to navigate the complex web of meanings embedded in words,” Gameiro says.
Language complexity can affect treatment and patient care. “Pain can only be transmitted through the ratiophor,” continues Urlaub. Smiley’s Face and 1-10 Scale – Pain Measurement Tools that English-speaking healthcare professionals may use to assess patients – they cannot move well across race, ethnic, cultural and linguistic boundaries.
“Science needs to have a mind.”
While LLMS could potentially help scientists improve healthcare, there are several systematic and educational challenges to consider. Science can focus on the consequences to eliminate those that it intends to help, Seri argues. “Science needs a mind,” he says. “There are patents that measure student effectiveness and miss points by counting the number of papers they publish.”
According to Urlaub, the point is to carefully investigate, citing what philosophers call epistemological humility, simultaneously acknowledging what we don’t know. Investigators argue that knowledge is tentative and always incomplete. Deeply held beliefs may require revision in light of new evidence.
“Nobody’s spiritual view of the world is perfect,” says Seri. “We need to create an environment where people can comfortably acknowledge their bias.”
“How do you share your concerns with language educators and others who are interested in AI?” asks urlaub. “How do you identify and investigate the relationship between health professionals and language educators who are interested in the potential of AI to help eliminate communication gaps between physicians and patients?”
In Gameiro’s estimation, language is more than just a tool for communication. “It reflects the dynamics of culture, identity and power,” he says. Misunderstanding can be dangerous in situations where patients may not comfortably explain pain or discomfort, as demanded by a culture of succumbing to people who are perceived as authority figures.
Change the conversation
Language-equipped AI facilities will help healthcare professionals navigate these areas more carefully and provide a digital framework that provides valuable cultural and linguistic contexts that allow patients and practitioners to rely on data-driven research support tools to improve dialogue. Institutions need to rethink how to educate healthcare professionals and invite communities to serve the conversation, the team says.
“We need to ask ourselves what we really want,” Seri says. “Why are we measuring what we measure?” The biases we bring to these interactions — doctors, patients, their families, and their communities — remain barriers to improving care, says Urlaub and Gameiro.
“We want to connect different ways of thinking and make AI work for everyone,” Gameiro continues. “Technology without purpose is a massive exclusion.”
“Collabs like this allow for deeper processing and better ideas,” says Urlaub.
Creating spaces where ideas about AI and healthcare can become action is a key component of your project. The Language/AI Incubator hosted its first Colloquium at MIT in May. It was led by Mena Ramos, physician and co-founder and CEO of the Global Ultrasound Institute.
Colloquium also featured presentations from Celi, Alfred Spector, visiting scholar in the Department of Electrical Engineering and Computer Science at MIT, and Douglas Jones, senior staff at the Human Language Technology Group at MIT Lincoln Laboratory. A second language/AI incubator colloquium is planned for August.
The greater integration between social science and hard science could increase the likelihood of developing viable solutions and reducing bias. By enabling changes in the way patients and physicians view the relationship, providing shared ownership of each of the interactions can help improve outcomes. Promoting these conversations with AI could potentially speed up the integration of these perspectives.
“Community supporters have a voice and should be included in these conversations,” Seri says. “AI and statistical modeling can’t gather all the data needed to deal with everyone who needs it.”
Community needs and improved educational opportunities and practices should be linked to an interdisciplinary approach to knowledge acquisition and transfer. The way people see things is limited by their perceptions and other factors. “Whose language do we model?” Gameiro asks about building LLM. “Which variety of speeches are included or excluded?” It is important to remember these when designing AI tools, as meaning and intent can shift across these contexts.
“AI is a chance to rewrite the rules.”
While collaboration has many possibilities, there are serious challenges to overcome, such as establishing and scaling technical tools to improve patient provider communication with AI, expanding opportunities for collaboration in non-alienabled communities, and rethinking and remodeling patient care.
But the team is not daunting.
Seri believes there is an opportunity to address the growing gap between people and practitioners while addressing the healthcare gap. “Our intention is to reclaim the strings that were cut between society and science,” he says. “We are able to allow scientists and the public to investigate the world together, and we acknowledge the limitations that have arisen in overcoming their prejudice.”
Gameiro is a passionate advocate for the AI’s ability to change everything you know about drugs. “I’m a doctor and I don’t think I’m hyperbolic when I say I think AI is an opportunity to rewrite the rules of what drugs can be and who we can reach,” he says.
“Education transforms humans from objects to subjects,” Urlaub argues, explaining the difference between indifferent observers and active and involved participants in the new model of care he wants to build. “We need to better understand the impact of technology on the boundaries between these states of existence.”
Celi, Gameiro, and Urlaub each defend a space like Mithic through healthcare. This is where innovation and collaboration takes place without the kind of benchmarks that arbitrary benchmarking agencies have previously used to mark success.
“AI will transform all these sectors,” Urlaub believes. “Mithic is a generous framework that allows us to embrace uncertainty and flexibility.”
“We want to employ our power to build a community among our diverse audience and acknowledge that there is no answer,” Seri says. “If we fail, that’s because we couldn’t dream big enough about what a reimagined world would look like.”
