"People have the wrong perception of AI": Q & A with Richard Kimera
Richard Kimera is a lecturer at Uganda’s Mbarara University of Science and Technology and part of the DASH in Africa organizing team. He also chairs the annual Innovations Tech Summit for Africa, a convening of public, private, and government agencies and research institutions aimed at sharing research innovations and opening channels of communication between government, universities, and the public. He currently coordinates activities of the Data science and AI research group at MUST, as well as the android Softech team.
Kate Raphael: What led you to work at the intersection of AI and health? Can you speak to your experience and how you got to where you are now?
Richard Kimera: AI and health is a relatively new area that I’ve been investigating, but I first became involved in it when I started coordinating a collaborative course at MIT. Each year, we’ve had different topics and themes for the course.
Previously, we were emphasizing computer systems for the developing world, and in 2018, we focused on using big data to build health systems in developing countries, and that’s how I first got interested in this field.
Recently, I’ve been trying to learn on my own and work with students as a lecturer. I’ve been using data science and AI courses, and now I’m working on building a model that can detect breast cancer and predict breast cancer recurrence.
What are the most pressing health issues that can be potentially addressed by AI and data science in Africa, and specifically where you work?
The most pressing issues revolve around emergencies and accidents. In Uganda, a lot of accidents are caused by motorcycles and taxis. However, there aren’t means of collecting and reporting this information in real time, so it’s hard to develop systems to address the problems. If we had the right information, then we would be able to use AI to predict how certain situations, weather, or behavior could lead to an accident.
Another major issue is about rape cases. Most interventions attempt to address issues after rape has occurred, but that means that someone has already been raped. It’s often very difficult for people to report rapes, but with AI, we have the potential to build solutions to fight stigma. We can utilize robotics and chat bots which allow people to share information with a system that won’t laugh at or stigmatize them. People can seek help and medication immediately in a more closed environment.
What challenges do you anticipate in trying to address these specific health problems with AI?
In the case of using AI models to predict breast cancer, there have been a number of challenges. If the same cancer reoccurs, families might not be able to raise the money to treat the same person a second time, and they may give up and let people die. But, if there is a way to detect earlier, we may have more time to detect and treat the disease. I started wondering what we could do to help raise funds, maybe community fundraising, or even car washing to raise the funds. So, I built a platform that was meant to help with crowdfunding and it was supposed to look at how locals can help with health financing, as well as connect innovators to where problems are and then provide solutions.
With regards to the issues of rape, we developed “Rape 0.” This technology tries to look at what can be done to try and prevent rape. For example, “What time should you walk and if you’re alone, what places should you be aware of?” We’re working on a mobile device that could have sensors and could be switched on or off, so if you sense if something is going to happen, you can switch it on. This device would be able to generate and send a message in the phone app that will in turn send a popup and the device will start recording and will start calling a few trusted friends. As all of this is happening, it will send that information to the server online.
An additional section looks at post-rape care. If you’ve been raped, then what next? We’re trying to provide information on how people can access medication and counseling for trauma. From there, we identified the need for chat bots to counsel rape victims. We are also looking at voice capabilities so that people can speak to chat bots. How can we translate the text back to voice rather than voice to text?
The biggest problem with that is that we have a language barrier: most people do not speak English and sometimes we have doctors that can’t speak the local language. In Uganda, we speak around seven languages, so we’re trying to bridge these gaps via a local language repository that will assist with translation.
What are the “right approaches” for integrating AI into clinical care? What could we be doing, or what aren’t we doing that we should be doing?
First, we need to train more personnel. We have very few people who are experienced in both AI and health care.
Second, we need to communicate to our medical workers that AI is not here to replace them. They’re scared of being replaced, but they need to know that AI is here to support and help them improve, not take their jobs.
The third thing pertains to the way we collect information. We need to change our systems so that what we collect is in a format that can easily be analyzed by AI and data science systems. Often, the technology does not exist to collect and store this information, or it’s collected and thrown away. We need to collect and store these data in the proper format.
Finally, we as academic institutions, together with the government, need to establish collaborations to show that this work isn’t just aimed at research, but improving community. People have the wrong perception about AI.
Who do we need to partner and collaborate with to better address health challenges using AI? Who is often at the table, and who do we need to engage who isn’t often in the conversation?
I would start with the people who collect information or have a one-on-one connection with patients. Patients give us critical information and if we get that wrong, then the whole system fails. So, we should start with medical workers and train them in how information should be collected.
Two, we need to have real discussions between computer scientists and medical practitioners because this is a field that combines both those stakeholders. If these two see each other as competitors, we can never make it work. We should bring in NGOs that work in health care; they spend most of their time reaching out to local people, and they understand how a community works.
Research institutions are critical, but some of our communities don’t fully understood the value of research. Right now, we have so many research groups here, but people haven’t really established this research. This is where international partners should come in so that we can learn from what has worked and what has not. Most of the strategies in developing countries have been tried in developed countries, but there are a lot of different variables. So, we aren’t just interested in reproducing, but in refining and developing and finding the solutions that work for us. International organizations also bring in funding, which we really need.
Are there additional projects you’re hoping to pursue?
I’ve been looking at tools used to view solutions, and that’s often from the programmer’s perspective. However, that perspective is limited so it’s hard for people to solve health care challenges; if we can find ways of utilizing AI that are fast and easy for our population to learn, work for laptops, and are cheaper and faster, then we can easily work on how AI can improve health care.
Second, I hope to utilize the internet of things to build health care solutions and help us collect information. For example, sensors can collect information on pollution or weather patterns, but the sensor is going to make mistakes, so I would wish to utilize technologies to better process data and develop solutions.
Finally, in the future we need to think broadly about diverse health problems and solutions. Usually in developing countries, outside countries thinks that in Uganda, the society is mostly built on agriculture, and that the biggest challenge is HIV. But the problem is that if we just focus on those, we neglect other issues that we need to address so they don’t become massive problems later. If we develop solutions in Uganda, outside partners don’t want to fund the potential solutions to pressing issues. It’s very hard to focus on this area if you don’t have support, but that’s where I’m hoping to focus.