Can AI be evil? - The ethics of Data Science and Artificial Intelligence

von Laura Braun (Studentin Data Science)

Are you thinking about dystopian science fiction movies where robots and artificial intelligence become super-intelligent, evil and destroy humanity?

I am sorry to disappoint a science fiction fan right now, but this article is not that spectacular. For one thing, AI isn’t really intelligent and for another, it’s pretty far away from ending up in a scenario you have seen on TV. But evil, also according to Hannah Arendt, is often more banal than expected. In this article, I am going to look at critical thinking about ethics in Data Science and AI and why a Data Scientist or an AI engineer should have a basic overview of it.

Let's start with ethics in the context of AI and Data Science

Ethics or moral philosophy attempts to resolve questions of morality by defining concepts such as good and evil or right and wrong. The fundamental question of this subfield is that of the "good life". In the context of AI ethics, we therefore need to discuss what technologies and systems our society want to use, how to develop them, what risks they pose, and how we can control them.

To get a deeper insight into this topic, it is necessary to analyze the ethics of Data Science and AI from different perspectives. There are a lot of interconnected dimensions that involve a large number of current debates. For this short article I have selected some debates, clustered them into two different dimensions and briefly discussed them.

The first dimension – liberty and privacy rights

Privacy & Surveillance

Since privacy debates no longer focus only on state surveillance, but also involve other agents as well as companies and even individuals, we are increasingly losing our right to privacy, such as the right to control information about oneself and the right to secrecy. Although regulations are in place, technological change and digitization have become too rapid and important players such as the “big 5” have become too powerful.

The result is certain anarchy in which people have lost control of their data and are exploited in a way that Shoshana Zuboff calls “surveillance capitalism”. Even by monitoring online behavior, it is possible to get insights into an individual’s mental state and manipulate them, which leads us to the next point.

Manipulation of Behavior

Manipulation of behavior not only means that people are influenced in their consumer behavior by targeted advertising on the internet, but can also have dangerous consequences for society. Social media has become the best place for political propaganda and political parties have discovered microtargeting for themselves. Manipulating people’s behavior or opinion in this way can pose a serious threat to our democracy. How Data Science and especially which methods can be used for Social Media Analytics can be read in another blog article. [LINK]

The second dimension – discriminatory bias

Bias in Decision Systems and algorithms

Automated AI decision support systems and predictive analytics use data to produce outputs that can range from the very trivial (movie recommendations) to the extremely important (organ donation). Then a problem may be that the data has a systematic error, such as a statistical bias.

Machine learning based on such data would then not only fail to detect the bias but would encode and automate the "real bias." This can lead to discrimination, such as at Amazon, where it was discovered that an automated hiring screen systematically discriminated against women, presumably because the company had a history of discriminating against women in hiring.

Another problem is that laypeople tend to think of data as inherently objective, forgetting that it can only be as objective as the people and processes behind generating and collecting it.

But don't worry, you'll learn to look at data with a Data Scientist's eye, not a layperson's, from the beginning of your studies.

Why should a Data Scientist or AI-Engineer care about ethics and why is it important to teach ethics also in data science courses?

Artificial intelligence plays a role in the lives of billions of people today. Sometimes unnoticed, but often with profound consequences, it has reached almost every single area of our lives and is changing our society.

Looking at the bigger picture, as a data scientist you have a very powerful job and with your work you will in some way impact people's daily lives. AI and Data Science are not only focused on business purposes, but can also be found in critical areas of life such as health, justice and education. Of course, it is first and foremost the responsibility of politicians, society, and humanities as well as social scientists to think about ethics and discuss what values and ideas our society have regarding AI and data science.

We already have laws like the GDPR to protect our data, and the EU has also drafted AI regulations. So for that reason alone, it is beneficial for a future career to think about and discuss the ethics behind the laws as well as the law in data science classes.

Of course, it's important to keep in mind that in your later job as a Data Scientist, you will likely have little to no opportunity to critically question whether the social media algorithm, policy model, or teacher evaluation is ethical. Nevertheless, I think it is a good approach to enable a new generation of Data Scientists to do so. Besides the validity of the data, the usefulness of the used methods, and the safety and reliability of the model, checking ethical standards should make it an integral part of data analysis and AI engineering.

Can AI be evil?

Returning to the question of the title, the answer should be clear.

AI cannot be evil.

Why? Because AI is technology, neither magical nor intelligent, just technology. The point is humans do Data Science and create AI. Humans can be evil, although, as Hannah Arendt noted, they are not even aware of it.

This is the reason why we need a critically thinking young generation of Data Scientists and AI engineers who are aware of the power they have and are also able to reflect on it ethically.

Finally, I need to annotate that the possibilities that Data Science and AI offer us to enrich our lives or to conduct new scientific research (e.g. innovations that address climate heating) are innumerable. So start your Data Science studies at WHZ, keep in mind the relevance of ethics for data analysis and AI engineering and co-create the future.

Sources:

Zurück