Prof. Sharit Kraus from Bar Ilan will receive a grant of 2.3 million euros for research in the field of artificial intelligence

The grant, which will be spread over five years, will be given to her by the European Research Council, with the aim of developing a system in which the computer will negotiate with human users

Prof. Sharit Kraus from the Department of Computer Science in the Faculty of Exact Sciences at Bar Ilan University recently won a grant from the European Research Council (ERC) in the amount of 2.3 million euros for a period of five years.

Prof. Kraus, who last week was awarded the AMT award in the field of computer science for 2010, won the Advanced Investigator Grant for research in the field of artificial intelligence. This is a relatively large grant for a single researcher, which will be used to develop a "computerized agent" - a smart computerized system that is capable of making decisions. Following these decisions, the agent software will be able to negotiate with human users, and in fact - argue with them to reach results.

"The applications of developing 'computers debating with humans' are multiple, starting from personal training of people in negotiations through smart e-commerce systems to applications in the field of training medical teams," said Prof. Kraus.

According to her, "most of the systems developed for the needs of negotiations with people have concentrated on the field of bargaining - bargaining - where both parties make offers until a decision is reached or not." She said that as part of the project, "the computers will be able to try to convince the person with the help of computational tools, such as: deciding what information to reveal, trying to change his goals and preferences."

According to Prof. Krauss, the "computerized agent" will be installed on a computer or robot and will help the elderly book an appointment for physical therapy. If the elderly person wishes to perform the physical therapy in the morning hours, when he is alert, while the system manages to schedule an appointment only in the afternoon, the elderly person may ask the software to try to change the time. If it does not succeed, the software will try to convince the elderly that the set date is reasonable. Among other things, she will suggest that he rest before the treatment and reinforce the importance of physical therapy. "Such a degree of artificial intelligence is the dream of my life," she said. According to her, "Persuasion is done by adding information and trying to provide a solution - something that is missing in today's agent applications."

Kraus prefers this application over e-commerce applications. However, she believes that achieving her goal is difficult, and she would be willing to settle for systems that train doctors and medical staff to interact with patients when the agent plays the patient.

"In order to succeed in the task, we must develop algorithms and methods for modeling people - preparing a computer model that will simulate the behavior of people according to characteristics common to groups of similar people," said Prof. Kraus. "Among other things, it is required to use insights from game theory and learning algorithms from computer science. So far we have succeeded, using models from the fields of economics and psychology, to improve the behavior of an agent who was able to more easily guess what people would choose. An example of this is when two people, who do not see each other, are offered to divide a sum of 100 dollars given to them in one dollar bills into two piles. If both people choose the same division - they get 100 dollars, and if they choose different divisions - they get nothing. For the computer, any division is equally possible."

According to Prof. Krauss, these are difficult research questions, and as mentioned - she hopes to make progress in these areas to improve the quality of our lives in the age of computerized agents.