Skip to 0 minutes and 9 secondsHello, Shaheen. Hello. Good to see you here. Thank you. You are a Ph.D. student doing your research on cholera modelling. When did you start your Ph.D? I started in July 2014, two years ago. So you are about half way, your research now? Yes. OK, what was your background before you started your Ph.D. My background is computer science. And also I did my MSc in Geoinformatics. Did you have any experience working with agent-based simulation models? Yes. During my MSc, I worked on agent-based modelling. I used it to simulate the spread of pertussis in the Enschede region. OK. Now, during your Ph.D, what you're trying to do is implement machine learning in an agent-based simulation model, in the cholera model.
Skip to 1 minute and 8 secondsWhy is it important to do that? Well, I'm using the learning algorithms, or methods, to improve the agent-based model I am using for my cholera model. So what you mean by improve is actually make the agents-- Smarter.
Skip to 1 minute and 29 secondsThe agents we have in the model, they are more abstract. So we need to make them smarter, can make a decision based on their learning experience. So what you would like is that your agents check if they think they are at risk and then, perhaps, change their behavior. Yeah, of course. During epidemics, normal people think that they are in risk of, so they try to protect themself in a way and make some decisions that effect on them and their, also, children, for example. So that's why we need to have smart, let's say, agents that these agents can learn from their experience, from other experience, and improve their decisions to protect themselves. It sounds like a very important.
Skip to 2 minutes and 22 secondsAnd it's actually on the edge between computer science, on the one hand, and agent-based modelling, but also behavioral science. Sure.
Skip to 2 minutes and 34 secondsHow do you collect information on the behavior of people? Or the change in behavior? Yeah, well, a researcher in the behavioral science gave us theories that explain reasoning why people react in a specific situation. But this is not enough since we, in agent based modeling, we need to validate our models. That's why we need more information to see the reaction of normal people. You know, as a human being, we never follow a specific rule, or a specific equation, let's say, in terms of math, that explains our behavior. That's why we need more to be in contact with real people, let's say, to have or register their reaction in such situations.
Skip to 3 minutes and 31 secondsSo I understand that you would like to ask the participants to actually-- Yeah, it will be very useful for me as a researcher, for my research, also, and also for further studies, like for policymakers, for people that know what to do in case of epidemics to prevent themselves and their children also. OK, so I understand you have prepared a small questionnaire that you would like to ask the participants. Yeah, of course, and I'm inviting people to join us and participating in this questionnaire. Their answers will be very valuable for us and to improving our model, too. Well, I hope that many of our course participants will do your questionnaire. And I wish you good luck with your Ph.D.
Skip to 4 minutes and 26 secondsThank you, so much. Thank you.
An agent is not the same as a human being. Humans have the ability to learn new behavior, and to change their behavior. To make the outcomes of our models even more useful, we try to let our agents behave more like human beings. These agents use artificial intelligence to decide which water to use. Shaheen Abdulkareem is one of people working on making smart agents. Shaheens tells a little about her research in the video. And she also invites you to participate in a questionnaire (which will follow in the next step).
In the cholera model agents make a choice on what type of water to use. When they can afford it, they will buy safe water. If not, they will use the river water.
You learned that the behavior of agents is captured in behavior rules that have a condition and a behavior part. If the condition it true, execute the behavior. When you construct your behavior like this, throughout the simulation the behavior will remain the same. Nothing that happens can stop agents from executing the “if… then” rules.
This is not how human beings behave. There are many reasons why people change their behavior.
Change of behavior
The behavior change can come from the person itself, but can also be the result of what the person “senses”. Perhaps there are many reports on cholera in the media and the fear of getting cholera increases to a level where the person will start to take precaution. Perhaps when a person in the direct surrounding gets ill, this really stresses the fact that you may be at risk too. The real person cannot see if water is free from cholera bacteria or not. This is why people may judge the quality of this water via what they can see. The fact that someone always uses the river water may give this person a sense of safety, as he/she did not get cholera before.
- The decision making process is complex
- It may involve a combination of different types of information
- The same information may lead to a different decision for different agents.
So to make the outcomes of our models even more useful, we try to let our agents behave more like human beings. These agents use artificial intelligence to decide which water to use. This is what Shaheen Abdulkareem is working on.
© University of Twente