Berkeley Dieter worster: At first, we thought that for those who don't like arithmetic, we could do experiments. Once people understand operation and see that they can also predict through operation, they will start to use this method more. But after the experiment began, we found that things were exactly the opposite of what we expected. People are happy to accept it when they don't see the process of operation and when they don't see the mistakes in operation. But in the experiment, if people see that the operation is wrong, they are unlikely to use and accept the operation.
Wharton Knowledge Online: It's interesting, because … most of the data are true. Why don't people accept it? For some reason, the operational data has not been accepted by people.
Macy: This is the real focus of Buckley's research, and his insight also guides us ... People are more likely to accept people's subjective judgments rather than the calculation results. No operation is foolproof. I mean, even very, very accurate calculations are not perfect. Small mistakes are practical problems that need to be overcome in the operation process.
Wharton Knowledge Online: Why? Have you analyzed it?
Dieter worster: I think one of the reasons why people don't want to apply the arithmetic method after seeing the arithmetic error is that once people see the arithmetic error, they will think that they may make mistakes again in the future. To some extent, this is also correct.
Messi: But one of the advantages of operation is its consistency.
Dieter worster: That's one of the real reasons why we like this way. It's coherent.
Messi: People mistakenly believe that human beings will not make mistakes continuously, and will even continue to improve and improve. But in many cases, this is not the case. Compared with computers, human beings make more mistakes.
Wharton Knowledge Online: Computers or arithmetic programs can adapt and adjust when facing new data or program changes, but it is difficult for human beings to do so.
Dieter worster: Yes, on the one hand, the calculation method can be constantly changed, and it can be continuously improved in the future. But on the other hand, people should also realize that it is impossible to make perfect predictions, because many results in the real world are determined by accidental factors. Therefore, even if your prediction is completely correct this time, it is impossible to guarantee that it will be correct every time.
Messi: Hilly Einhorn, a researcher at the University of Chicago, wrote a very reasonable sentence in his paper, "Accept mistakes so as not to make more mistakes". You must accept that your model is not perfect. Admitting this will help you make fewer mistakes, because human beings may be more imperfect.
Wharton Knowledge Online: Yes.
Messi: We realize, and we also have evidence that people always pursue perfect predictions and are unwilling to give up the opportunity to be right every time, even if it is impossible.
Wharton Knowledge Online: But your research also mentioned that there are economic considerations.
Messi: We have a strong inference, but the evidence is not sufficient at present. We infer, ironically, that the more important things are, the more people reject computing. If they don't care much about the result, they are more willing to accept the help of the computer. For example, the Super Bowl in America. When it comes to competitions, people will think, God, we can't let arithmetic decide the outcome. At present, we have some similar evidence, but we haven't studied this problem systematically.
Dieter worster: At present, we are carrying out this research in the laboratory.
Wharton Knowledge Online: Sometimes the data results may be completely contrary to your prediction, which is obviously somewhat surprising. Please share your understanding of this situation with us.
Messi: When we first encountered this situation, we obviously had to copy the data to ensure that the same results would be reproduced when we repeated the experiment. Then we will look for other reasons that can reasonably explain this situation. For example, we do find that participants lose confidence in the operation after seeing the operation process. People who have never seen the operation process may be more confident. We began to check all the data, and all kinds of indications show that it is because we have seen the operation process that people are reluctant to accept the operation method.
Dieter worster: This kind of confidence data is very interesting. Even if people make mistakes, they will not lose confidence in themselves. When you make mistakes, you are still confident in yourself, even though people may make more mistakes in experiments than machines.
Wharton Knowledge Online: What do you do with the data of this study and how do you apply it to the real world?
Messi: In our current project, we are exploring how to persuade people to use computing tools, even if they know that machines can make mistakes. We have conducted some experiments, some people use models to predict, some people rely on their own prediction, and some people can adjust the prediction results of the model with their own judgment, but there are conditions to do so.
for example, the data operation gives a result, and you can adjust it up or down by 5 points. I found that people like this way, and prefer to integrate personal factors. In fact, even if something goes wrong in this way, people know that something is wrong, but they don't necessarily lose confidence in this way. Therefore, as long as people can participate in decision-making and use their own judgment, they will be more willing to use computing tools.
Wharton Knowledge Online: The Boston Globe published an interesting article introducing your research. The article mentions an interesting case in a Norwegian restaurant. What I want to say is that the restaurant doesn't let the sommelier tell you what is the best wine, but explains it through a tablet computer.
Macy's: Why do you think people are more likely to accept the wine recommended by data calculation than the wine recommended by the bartender?
Dieter worster: I guess they are reluctant to accept it.
Messi: You don't think it will be accepted. Why?
dieter worster: yes. If in your field, people believe that human beings have special insight, but this is beyond the understanding of machines, then people are probably reluctant to use data operations. I think so, but obviously we haven't confirmed this. Computers have never tasted wine, so I don't think people will believe the results given by computers.
By the way, I guess computers actually "taste wine" ... People certainly don't want to hear this, but I guess it will.
Macy: If the computer hasn't tasted wine, it may in the future.
Dieter worster: We will figure it out in the future.
Wharton Knowledge Online: It's like opening a door to different fields of exploration, showing people that data can affect all aspects of real life.
dieter worster: indeed. Because of this, we felt that this research was very important at first, because the applicability of this project is increasing. With the development of big data, more and more people try to help them make decisions in various fields with the help of data operations. We need to know more about what can help people overcome the obstacles of not believing in machines.
One of our motivations in this work is to deeply understand how deeply people's prejudice against machines is, how much people need it, and how hard it is to change this judgment.
Wharton Knowledge Online: Obviously, prejudice is a common problem in daily life.