When Algorithms Manipulate Human Beings

Yolande

Most of the people living in our age have heard of algorithms. When we open WeChat, we often see advertisements targeted to send, buy books in online stores, and have a series of books that are recommended to you later. This is a data era. The age of big data is also an algorithm era.

The author of the algorithm controlled life is a math professor from England and Sweden, who knows that I am a math student and I am engaged in financial science and technology research, so please help me look at it. Mathematicians write a science book, which is a selling point in itself. I remember that in addition to mathematics textbooks, I rarely read bestsellers written by mathematics professors. There are many interesting examples in this book. Some ideas deserve our consideration.

Personal data involves personal privacy, and analysis of personal data will also reveal personal privacy: is it lawful for consumer portraits and ads to be targeted? I think it depends on what effect the personal data analysis can achieve. Such algorithms generally use principal component analysis and regression analysis. These two methods are not very sophisticated themselves. Many social networking sites are used, such as Facebook and twitter, which are mainly used to classify users. All kinds of activities that we go online everyday belong to personal behavior data, leaving a record on the website, which reflects our state of life. For example, the content, praise and information sharing are related to personal behavior, opinions, preferences, intelligence quotient and personality. Using algorithms and accumulated mass data, we can examine everyone from the most sober and rational perspective.

Our brains also form opinions on others, but the brain can only process up to 3 dimensions, and computers rely on algorithms to categorize each person quickly in hundreds of dimensions. In theory, human behavior data can be used to track every emotion and predict in terms of consumption choices, interpersonal relationships and job opportunities. The more data, the more reliable the classification. There are no clear rules in the social network for collecting and analyzing users' data, which is controversial at the ethical level. A case of concern is the Cambridge analysis company manipulating election events. The company is accused of influencing voters by using network directed marketing and small scale polls. In July 2019, Facebook paid a $5 billion fine on the matter, the largest privacy violation penalty in the world so far.

There is no difference between Market Research and the methods adopted by data analysis firms and researchers decades ago, that is, using well-designed algorithms to calculate sorting or probability in each dimension, but the amount of data used is much larger now, or even the use of big data. Under such a scale, people have to use computers and algorithms to analyze. For example, face book regression algorithm for nearly 20 thousand people's point of praise data analysis, in 9 times 8 times successfully predicted the personal political position, accuracy and reliability is very high. For example, a person who likes Lady Gaga, Starbucks and country music is more likely to be a Republican. In order to win support, the Republicans should focus on the people who drink Starbucks coffee.

Collect user's social network data, and customize the content that meets the specific audience's viewpoints and preferences according to the personality characteristics of the target population, and then guide the target population to change their behavior. If the algorithm develops to this level, it will obviously infringe on the scope of personal rights. Targeted delivery of information, including advertising, may be beneficial to a particular audience, or it may make the other person feel unsafe or even offended. When the algorithm makes use of personal data to predict the accuracy of personal behavior and preferences up to 85%, whether the use of personal data is reasonable or legitimate becomes a prominent problem. I personally think that the business model based on search algorithm must first obtain personal approval.

Data algorithms are mostly "black boxes" for most people. In fact, there are similar problems in the stage of AI development. For instance, deep learning is likely to inadvertently introduce all kinds of discrimination problems existing in human society into the algorithm. Algorithm discrimination will make social gap bigger and seriously affect social equity. Many examples of algorithmic defects are given in the book. Taking electoral forecasting as an example, the manual sampling survey is closer to the actual result than the algorithm. The Gallup poll's error in the US presidential election has been decreasing from 1940s to 2020, and now it is basically below 5%. The worst predictor of the algorithm is similar to that of chimpanzee throwing dart, and the best is only about 60%. Data algorithms are also being exploited, such as book sales, which have improved sales by using some of the top priority of search optimization algorithms, but the result may be a lower reader feedback rating. The same method can improve the citation of academic papers. Taking this index as the sole criterion for assessing academic achievements, it will inevitably lead to the reverse incentive and Reverse Elimination in academic circles. We should understand the possible problems and avoid being misled or manipulated.

?