We Humans and the Intelligent Machines
The use of algorithms has long since ceased to be science fiction and has become reality. Seemingly intelligent machines are part of our lives. They surround us and enter ever new, sensitive areas of life. They analyze human behavior and shape modern societies. We should deal with them before we get used to them. We must therefore re-evaluate the relationship between humans and machines. How does artificial intelligence affect us, our lives and our society? Where can algorithms enrich us, where must we put an end to their threatening omnipotence? Answers to these pressing questions are sought in the book “We Humans and the Intelligent Machines” (“Wir und die intelligenten Maschinen”) by Jörg Dräger and Ralph Müller-Eiselt, which was published in 2019 and is now also available in the English language edition.
Defeating cancer before it develops. Stopping crime before it happens. Getting the dream job without the right connections. Serving justice freed from subconscious prejudices. All of that sounds auspicious, yet the negative narrative is just as impressive: healthcare systems which are no longer based on social solidarity, minority groups which suddenly find themselves disadvantaged, individuals who are completely excluded from the job market. In this scenario, people become playthings, the victims of digitally determined probabilities. Whether promise or peril – the changes will be radical.
In their book “We Humans and the Intelligent Machines”, Jörg Dräger and Ralph Müller-Eiselt illustrate the scope of algorithms and demonstrate their social relevance in more than 40 vivid and surprising case studies. The book is neither dystopia nor utopia. It illustrates the opportunities for a better society without losing sight of the risks of algorithms.
We must re-evaluate the relationship between humans and machines!
We humans are not perfect: too much information overwhelms us, we make inconsistent decisions and we discriminate unconsciously. As “augmented intelligence”, algorithms help us to compensate for human weaknesses. The authors consciously advocate this choice of words, since the term “artificial intelligence” suggests that human and machine intelligence are virtually the same thing. And it suggests that artificial intelligence can replace human intelligence just as an artificial hip joint can replace a natural one. Since algorithms make faster, more reliable and better decisions than humans anyway, they will displace us – so the fear goes. It seems as if men and machine were hostile to each other.
But this is wrong. However precise algorithms may analyze and assist, they would be lost without us. Humans have to define the goals of software and check it for errors. We are creative and reflective, we forge intrinsically motivated plans, we have social conventions, norms and ethical principles. Machines, even if they are called intelligent, are not and do not have all these things. Their intelligence is limited, they learn and operate within human limits; they do not have their own ethical compass. This results in mutual dependency: machines cannot do without us, but without them neither would we be able to cope with many things in the data age, or at least not so well.
That is why we need a new term. Augmented intelligence would be the better word. Because that is what it is all about: machines that only work on behalf of humans and selectively expand our capabilities. How this model of men with machine can succeed is shown in the book, amongst others, by the examples from medicine, police work and recruiting.
More than 40 case studies demonstrate the range and social relevance of algorithms
Doctors already benefit from algorithmically enhanced capabilities. Self-learning software can often diagnose faster, in some cases more precisely. It evaluates imaging procedures such as MRTs and CTs, typical tasks of a radiologist, particularly well. Computers are better able to detect tumors – but that does not make physicians redundant. It is only when doctor and algorithm work together that the error rate drops to an absolute minimum. In addition, this gives physicians more time for patient care and therapy.
The police also benefit from the analysis of large amounts of data. Patterns of past crimes help them to predict and fight future crimes. So-called predictive policing algorithms calculate in which districts or streets burglaries are to be expected, for example. Using this information, the police can better deploy their scarce resources to protect citizens. It is more likely to be where it is needed, rather than waiting for a crime to be solved. Augmented intelligence helps the police to carry out effective prevention work.
Algorithms can also make life easier for people in the labour market. For example, they help HR professionals to find the right candidates for advertised positions. Increasingly, an algorithm is being used for pre-selection – and can not only handle large numbers of applications, but also make suggestions without prejudice or discrimination. The final decision is again made by human professionals. They, in turn, have more time for the interview and can thus better assess whether a qualified candidate fits into the team. The interaction between human beings and algorithm increases the probability of finding the most suitable job match.
To put algorithms at the service of society, concrete solutions are needed!
The authors are sure: “Coding is political! It is up to us to use algorithms for what is socially meaningful instead of what is technically possible.” And: “We can use them to make our society fairer, more efficient and more humane. Achieving this is the political task of our time. “
To put algorithms at the service of society, they propose four concrete solutions:
- First, a broad societal debate. We as society must discuss where we use algorithms, for what purpose and according to what rules. Not all algorithmic systems are equally relevant to our society. But where intelligent machines determine our lives, there must be a public debate about their goals, their design and their effect.
- Secondly, effective control. Algorithmic decisions must be clearly recognizable as such and be verifiable, comprehensible and contestable in case of doubt by users or independent third parties. This does not require a standardized technical inspection for every single algorithm or even a separate algorithmic law. Instead, the legislator should above all supplement existing laws and strengthen long-established control institutions such as financial supervision or drug control authorities.
- Third, diversity and competition. Monopolies are harmful, as everywhere else, when algorithms are used. Only a diversity of algorithmic systems and goals can adequately represent social plurality, avoid discrimination and promote innovation. Essential for this is better access to the fuel of algorithms, the data. Only those who have access to high-quality data can succeed in becoming serious competitors on the market.
- Fourth: Algorithmic competency at all levels. Every citizen must be able to assess whether and how the decisions of an algorithm are relevant for him or her. Anyone who commissions, develops and uses software must consider its social consequences and ethical aspects. And we need a public sector that is strong in AI matters, one that regulates intelligently on the one hand and uses algorithms to foster the common good on the other hand. A state agency for algorithmic competencies would help to quickly build up the urgently needed abilities for this throughout the public administration.
The authors do not see Europe’s current backlog as an obstacle, but rather as an opportunity. “We don’t have to repeat the mistakes of others and can go our own European way, which should not see values and competitiveness as opposites. This European path gives the common good a higher priority than in the US and, unlike in China, preserves a high degree of individual freedom.”