How does working with AI impact on research?

‘The hysteria is quite uncalled for.‘ The calm body language which accompanies Jens Schröter’s words emphasises their significance – and stands in stark contrast to some of the statements made by other experts. The 55-year-old is Professor of Media Cultural Studies at the University of Bonn and is intensively involved with artificial intelligence (AI), a topic that has been causing a constant stir for several years now.

What interests him most is how other researchers deal with AI. Together with Prof Dr Anna Echterhölter from the University of Vienna, Dr Andreas Sudmann from the University of Bonn and Prof Dr Alexander Waibel from the Karlsruhe Institute of Technology, he is investigating the transformative impact of using AI in academic research. How is artificial intelligence changing science? Research in the era of learning algorithms is the name of the project funded by the Volkswagen Foundation and led by Schröter together with his co-applicants.

With his curly grey hair, well-groomed full beard, in which some white is now visible, rimless glasses and cosy-looking grey woollen jumper, Schröter would not stand out even among a group of avante guard artists. ‘I actually thought for a long time about whether I should study art or science,’ says the Darmstadt-born artist. Cultural studies gave him the interesting opportunity to ‘have a bit of both.’ He eventually enrolled for theatre, film and television studies, art history and philosophy at Ruhr University Bochum. ‘That turned a lot of what I thought I knew on its head – and I realised that I didn't really know a lot of things I thought I knew.’.

A universalist of media studies

Anyone who talks to Schröter today quickly realises that he is now at home in many different disciplines. ‘Some people say I'm the last universalist of media studies,’ says Schröter with a grin. His topics range from photography and post-monetary forms of economic organisation to three-dimensional images such as holography. Or in his own words: ‘I have a penchant for the unusual, and I simply enjoy doing experimental projects.’

‘What I really find interesting is that we no longer know exactly what is going on in machine learning,’ emphasises Schröter. Explainable AI, i.e. AI that makes it possible to understand how it arrives at its results, is therefore also a major topic today. ‘AI can also produce nonsense, because it's just statistics,’ says the researcher. But disciplines such as particle physics have been working statistics for a long time. ‘You probably have fewer problems with AI there than in some other sciences,’ surmises Schröter. But that is precisely one of the questions being investigated in the project.

The special perspective of media studies

But why of all people is a media scientist working on this? ‘In its megalomaniac moments, media science assumes that there is nothing that is not mediated by the media,’ jokes Schröter. Animals are also intelligent, can use tools, recognise themselves in the mirror – and ants have a highly developed division of labour and even practice agriculture. ‘But no animal is capable of building libraries, which is why animals will never develop beyond their present organisational state.’ Civilisation functions by writing down knowledge and passing it on. This ultimately means that every civilisation depends on its media. In addition, media science sees humans as an effect that is produced by media order: ‘Humans are a system consisting of body, brain, speech and stored code such as address, bank details, etc.,’ explains the researcher. What we now see is simply a new medium that can be used to analyse large amounts of data and also generate new data.

For Schröter, it is clear that there is no such thing as an AI effect on research. ‘Research is far too variegated to allow such a generalisation,’ emphasises the 55-year-old. Of course, AI can make data-intensive research more efficient or even possible in the first place. Other areas, such as art history, often do not require such a tool. Still others, such as climate research, have been working with complex computer simulations for a long time in order to tap into the complexity of their specialised field. Schröter can imagine that AI will be slow to find its way into this field, as well-functioning methods already exist.

Looking over the shoulder of AI users

The project team is approaching these questions from three directions. Firstly, the researchers are looking at AI from a scientific-historical perspective: ‘Data is not simply a given, but has to be measured and collected,’ explains Schröter. ‘Our focus lies on the people who don't appear in the data, how data sets are biased or where there are copyright infringements.’ Why is data the way it is and what are the consequences?

Secondly, the team looks over the shoulders of researchers: what software is used, what processes are utilised? Are existing frameworks used – or do the researchers build their own software systems? ‘We are trying to get a step-by-step picture of what it means for the respective discipline to work with AI at the present time,’ summarises Schröter. ‘What are the discussions, problems and opportunities that arise?’ The overarching question is whether AI is more likely to expand existing methods or even enable completely new avenues of research – perhaps even of a qualitative nature. Owing to the lack of traceability, AI changes how source data and results are linked and can therefore fundamentally change the relationship between researchers and their data

The challenge with this so-called ethnographic research is always to gain the trust of other people so that they can be completely frank with us. Another critical question is whether people actually behave as they would if the media researchers were not there. However, Schröter can already report a hardly surprising result: ‘There is never enough computing power, never enough money.’

Researching with AI about AI

Thirdly, the project team itself uses AI: ‘We – especially Mr Sudmann and Mr Retkowski – are working on systems with which we can analyse data for AI research,’ says Schröter. Researching with AI about AI, so to speak. This could result in considerable savings in time: ‘A major challenge in our project, though, is the rapid pace of developments,’ explains the researcher. ‘Just when we've begun to understand something, it's already become old hat.’

This office with its lack of distractions in the middle of Bonn's bustling city centre, the neighbouring computing machines that encourage you to look for parallels in the past – all of this seems to symbolise the attitude of media researcher Schröter. Whether young people will fall in love with AI bots and how autonomous cars will change mobility is something he finds exciting. But for Schröter, the biggest hype surrounding AI, the concern that it could destroy masses of jobs, is ‘just another episode of the weaver riots.’ From the media scientist's point of view, however, research is less likely to be about jobs: ‘Developing new ideas has so far not been the strength of AI, even though there have already been some astonishing results in individual cases.’

Technical revolutions do not change the fundamentals of life

With regard to the current hype, Schröter refers to history: ‘Every time there a new technology appears, some people shout that paradise is dawning, and others that the world is coming to an end. That was the case with radio, television, the internet and now AI.’ Technical revolutions are unlikely to fundamentally change the world: ‘Tomorrow we will still love, pay taxes and unfortunately also have capitalism.’

And so Schröter's interim conclusion after three years of project work can be applied not only to research, but also to society: ‘There is no AI revolution. And surely, that is in itself a remarkable result.’