We Humans and the Intelligent Machines. Jörg Dräger
Чтение книги онлайн.

Читать онлайн книгу We Humans and the Intelligent Machines - Jörg Dräger страница 12

СКАЧАТЬ the company (see Chapter 12). When asked what criteria the program used, the HR director replied, “I don’t know why this works. I just know it works.”11 Such answers forestall any debate about which candidates are rejected and why and whether there might be a systematic bias.

      A second example is provided by Germany’s Federal Ministry of the Interior. It used facial recognition software in a pilot project at the Südkreuz train station in Berlin to search for criminals and terrorists. Its official statement for the project reads: “We achieved a 70-percent and above recognition rate of the test subjects – a very good figure.”12 This means that the software correctly recognized seven out of ten wanted persons. But that is not the entire story. The ministry did initially not disclose the number of innocent passers-by falsely identified by the system. Its complete interim report has been kept under lock and key.13

      Both users, Xerox Services and the Ministry of the Interior, are thus making it more difficult to have a public discussion on the use of algorithms, one that is sorely needed. Both the question of possible discrimination in selecting employees and the right balance between surveillance and security needs are sensitive issues in a free society. Citizens can legitimately demand that users of algorithms assume responsibility and not hide behind a machine. More facts and figures need to be on the table for a real debate to take place. After all, only those who understand how their systems work can detect and eliminate errors and biases.

      Not only do we need effective algorithms, algorithms need us, too. We must therefore act in a way that is both competent and ethically responsible. In addition to the technical challenge, there are moral and legal aspects which must be addressed. Where seemingly intelligent machines judge people and errors quickly have a resounding impact, people must be able to discuss and define the goals machines are used for and comprehend their basic functioning at all times. We have a social responsibility to ensure that the software that governs our lives functions properly, that it is corrected when necessary, and that it receives the feedback it needs to improve. In cases where this is not possible and society’s key principles, including social solidarity, become endangered, we must not shrink from prohibiting the use of algorithms. In a democracy, artificially dumbing down artificial intelligence is a legitimate response (see Chapters 14 and 15).

       What algorithms can do for us

      A world without algorithms is hardly imaginable today. They have crept almost imperceptibly into our lives. Intelligent machines are now used almost everywhere that information is available electronically. The following nine chapters show the extent to which they are deployed and the impact they have. This second part uses practical examples to show how algorithms can make life better and more just for each of us and for society as a whole. Yet people and machines do not always complement each other in a meaningful way. Their interaction can also have negative consequences for individuals and society – be it unintentionally or by malice aforethought.

       An algorithm for algorithms

      It is precisely this tension that interests us. We want to examine those algorithmic systems that influence whether people can participate in society. For better or for worse. With consequences that concern us all, because they bring either social progress or serious disadvantages. Not all algorithms are truly relevant to society. Neither the spell checker in word processing software nor a car rental company’s computer-driven fleet management system will shake the foundations of communal life. They do not need public discourse – which, on the other hand, is indispensable if algorithms are to have a say in asylum procedures or prison sentences.

      To select the examples in the following chapters we used an “algorithm for algorithms” developed by researchers Kilian Vieth and Ben Wagner (see Chapter 14), which measures the relevance to society as a whole. Its main criteria: Are people being evaluated by the algorithmic system? How dependent are they on the result? How much political and economic power does the organization using the algorithm have? What is the system’s scope? The answers are assigned cumulative scores. The higher the overall result, the more relevant a system is to social participation. And, consequently, the more attention it deserves – not only in this book.

       Creating order in the jungle of algorithms

      The second part of this book dives into the world of practical applications. It looks at the impacts that algorithms can have on individuals and on society as a whole. We place more emphasis on clarity than on technical detail: Our focus is on the effect of the algorithms, not on their programming code. At the same time, we try to bring order to the jungle of different applications. The following nine chapters identify four impacts on the individual, four on society as a whole and one on our social interactions.

      At the individual level (Chapters 5 to 8), algorithms can meet personal needs more effectively, provide fairer access to goods essential for social participation, expand human capabilities and create space for activities we are particularly good at or like. Possible downsides are manipulation, exclusion of weaker individuals, an algorithmic arms race and reckless efforts to achieve ever higher productivity.

      At the societal level (Chapters 9 to 12), algorithms offer the potential to monitor the use of government services more accurately, distribute limited resources more efficiently, establish effective preventive measures for healthier and safer communities, and make fairer decisions. This is countered by risks such as excessive state intervention, the misappropriation of software, the weakening of social solidarity, and growing social inequality and discrimination. Last but not least, algorithms also influence personal relationships and our communication and values. They can strengthen cohesion, but also promote social polarization (Chapter 13).

       5Personalization: Suitable for everyone

       “No one can get outside his own individuality.” 1

      Arthur Schopenhauer, philosopher

      (1778–1860)

      Felix is unique.2 For six years, the 10-year-old from California has been continuously sending data into the cloud: pulse, stress level, exercise activity, blood sugar. Every day, tens of thousands of data points are collected by his various devices. Felix is probably the best-measured diabetes patient in the world. He is probably also one of the children whose diabetes is best managed. After all, a computer permanently evaluates all the data it receives from the boy’s smartwatch and his blood sugar monitor. An algorithm uses the information to calculate a therapy tailored to his exact needs. His parents constantly receive precise updates as to when Felix needs a snack or a dose of insulin.

      For Felix, this adds directly to his quality of life. He has Type 1 diabetes, an incurable autoimmune disease in which the body’s insulin-producing cells are destroyed. This causes him to oscillate between two states: hypoglycemia, which makes him restless and unfocused, and hyperglycemia, which makes him tired, weak and listless. To ensure Felix experiences these states as rarely as possible, his blood sugar level must be kept stable, providing the body with the СКАЧАТЬ