1.1.The real estate crisis, the collapse of large financial institutions, the rising unemployment, and the mathematicians behind the scenes who applied magic formulas became the leaders of these disasters.
1.2.Mathematics gradually stopped focusing on the dynamics of global financial markets and began to *** humanity itself.
1.3.Mathematicians and statisticians have been studying our desires, actions, and spending power, all the time ** our credit, and using the results to evaluate our performance as students, staff, lovers, and whether we have the potential to become criminals.
1.4.Some choices are undoubtedly well-intentioned, but there are also many models that weave into software systems that are increasingly controlling our lives to a greater extent.
2.1.Impact's Teacher Assessment Tool.
2.2.Characteristics: Opaque, not subject to questioning, inexplicable, and screened, positioned or "optimized" for a certain size of the public
2.3.It is used to evaluate the effectiveness of mathematics teaching and language skills teaching.
2.3.1.An attempt was made to evaluate a teacher's teaching with the help of an analysis of the test scores of twenty or thirty students.
2.3.2.It is also statistically unreliable, even ridiculous.
2.3.2.1.The sample size is too small, and everything can go wrong.
2.4.Rather than exploring the truth about poor teaching quality, the evaluation model does nothing more than visualize the problem with scores.
2.5.Sara Vessoki.
2.5.1.A good teacher recognized by the public.
2.6.The algorithm gave a rating weight that was half of her final rating, more than the school leaders and the community's evaluations.
2.6.1.Scores speak for themselves.
2.6.2.Scores are more fair.
2.6.3.Give her a bad review.
2.6.4.Washington had no choice but to fire her, along with another 205 teachers with an impact score below the minimum.
2.6.4.1.It doesn't exactly look like a form of politics or fractional determinism.
2.6.4.2.If the grading system determines that these teachers are unqualified, then others will perceive them as unqualified.
2.6.4.3.Many people, including the principal, vouched for a good teacher, and she quickly landed at a school in an affluent neighborhood of Northern Virginia.
2.7.Because of a model whose legitimacy and accuracy are highly questionable, poor schools lose a good teacher, while rich schools that don't fire teachers based on student test scores gain a good teacher.
3.1.Used to assess job applicants.
3.2.The idea of employers is that people who pay their bills on time are more likely to show up on time and follow the rules.
3.2.1.How many good employees they miss out on because they only focus on credit scores.
3.3.Unemployment leads to poverty, which in turn further lowers their credit scores and makes it even more difficult for them to find a job.
3.3.1.It's a vicious circle.
4.1.Weapons of Math Destruction, abbreviated as WMD
4.2.Trying to generalize human behavior, performance, and potential into an algorithm or model is not an easy task.
4.2.1.Mobilize fellow mathematicians against the use of sloppy statistics and biased models, which can lead to a vicious cycle.
4.2.2.Most mathematical** methods confuse the results with the actual situation, which ultimately leads to a vicious circle rather than a problem solving.
4.3.Statistical systems need feedback paths to ensure that operators are aware of errors in the system.
4.3.1.Big data companies like Google.
4.3.1.1.Researchers are constantly testing and monitoring thousands of variables. They can change the font of any ad from blue to red, serve different versions to 10 million users, and then track which version gets the highest click-through rate, fine-tuning the algorithm and actions based on user feedback as they go.
4.3.1.2.But Google's testing method can be said to be an effective use of data.
4.3.2.Amazon, Inc., needs to keep tweaking its model until the algorithm for user relevance recommendations is working properly.
4.4.Inherently flawed mathematical models are controlling the economy at the micro level, with implications ranging from advertising to prison operations.
4.4.1.Data scientists might say that no mathematical model is perfect, and those who suffer are collateral losses.
4.4.2.Think more about these amazing achievements made possible by algorithms, and ignore the imperfections.
4.5.There are many harmful assumptions about the construction of mathematical lethality**, and there is no disputing that these models are cloaked in mathematical precision, popular in the market, and put into use without testing.
4.5.1.A lot of mathematization** relies on its own built-in logic to define the situation it handles, and then justifies its output with its own definition.
4.5.1.1.This model is constantly self-reinforcing and self-developing, and it is extremely destructive.
4.5.1.1.1.And it is very common in our daily lives.4.5.1.2.The result is often a greater preference for punishing the poor.
4.5.1.2.1.Part of the reason is that mathematical models are designed to assess large numbers of people.4.5.2.Mathematically** excels at processing large amounts of data at a low cost of processing, which is also their advantage.
4.5.2.1.The wealthy often benefit from personal input.
4.5.2.1.1.High-end law firms or college-preparatory schools rely more on referrals and face-to-face conversations than fast-food chains or cash-strapped urban public high schools.4.5.2.2.The privileged class deals more with concrete people, while the masses are manipulated by machines.
4.6.Algorithms are like God, and the judgment of mathematical lethality is God's dictatorship.
4.6.1.Math Lethality** is like a black box whose contents are tightly guarded company secrets.
4.6.2.Maintaining the confidentiality of the algorithm also serves another purpose: if the people being evaluated are kept in the dark, they will be less likely to find vulnerabilities in the system.
4.6.2.1.They can only work hard, follow the rules, pray that the model is recorded, and reward their efforts.
4.6.2.2.There's no way to know exactly how the model works, which means it's hard to question or question the score given by the model.
4.6.2.2.1.How can you guarantee the legitimacy of the assessment if you yourself cannot explain the basis of the evaluation criteria?4.6.3.An algorithm is used to process large amounts of data, and based on the results, it suggests the possibility that someone could be a bad employee, a risky borrower, a terrorist, or a bad teacher, and that the score could destroy a person's life.
4.6.4.Mathematics is inevitably biased, miscategorizing some groups of people for a period of time, depriving them of the opportunity to find a job or buy a home.
4.6.4.1.The mathematical model operator does not think about these possible mistakes. The feedback they value is money, which is their fundamental motivation.
4.6.4.2.The model is designed to absorb more data, fine-tune the analysis results, and let more hot money pour in.
4.6.4.2.1.Investors reaped the benefits and decided to continue to invest more money in mathematical model development companies.4.7.You can't sue a math lethal**.
4.7.1.This is one of the reasons why we say that mathematical lethality is extremely destructive.
4.7.2.The model does not listen, does not give in, and is deaf to **, threats and cajoling, and logic, even when the person being evaluated has good reason to suspect that the data from which the conclusions were drawn is tainted.
4.7.3.If the automation system makes a mistake that is too obvious or something overall, the programmer does go back and change the algorithm.
4.7.3.1.Most of the time, the verdict of the procedure is beyond doubt, and the person operating the procedure can only shrug their shoulders as if to say, "Hey, what can you do?"
4.7.4.Victims of mathematical lethality** face a higher standard of rebuttal evidence than the algorithm sets for itself.