Thema1. Alexandra Brand.jpg

Between hope and hype

‘Artificial intelligence is a hype, a buzzword that we should not blindly pursue.’ But also: ‘AI will deepen my research and bring it a step further than it is now.’ There is a wide range of opinions. Nevertheless, four out of five Dutch scientists foresee that AI will have a considerable impact on society. Two thirds also believe that AI will have a radical impact on science.

Ai en de wetenschap. Diagram2.PNG
Text: Joop Daggers, image: Alexandra Brand

The figures come from a survey that the editorial board of Research commissioned among researchers in various disciplines at Dutch knowledge institutions. At the basis of this lies the opinion that was expressed in both the national AI strategy and the AI research agenda published in the autumn of last year (2019): AI has an impact on all science domains. In the words of the Dutch government: ‘AI is reforming how research is done in all scientific disciplines, from medical diagnostics (cancer research) and digital humanities (behavioural research) to astronomy (analysis of telescope data from black holes).’ Input from all those different disciplines is needed if we are to realise what the government calls for in its Strategic Action Plan for Artificial Intelligence: good solutions for societal issues in the areas of ageing, climate change, food security, healthcare and health, for example.

Opportunities, but also question marks

Therefore, AI will have an impact on all domains but also requires input from all domains. However, is that what all of the sectors and disciplines experience? In the private sector, the answer is yes. In a survey conducted in 2018 by EY and Microsoft that the government cited in its AI vision, 86 percent of companies in the Netherlands state that AI has a large impact on their sector. For science, the score is slightly lower: our survey among almost 1500 researchers reveals that two thirds (strongly) agree with the proposition that AI will have a considerable impact on science. Respondents from medicine, philosophy and computer science are, on average, the most outspoken with a score of 75 percent. Mathematicians (48 percent), lawyers (57 percent) and engineering researchers (61 percent) are more cautious. An even greater proportion of 82 percent of the researchers sees good opportunities for AI within their own discipline. Of the disciplines investigated (see box on page 7) historians and, interestingly, mathematicians, see those opportunities the least: on a scale of 1 to 5 they scored 3.4 and 3.7 respectively. The disciplines of computer science (4.6), medicine and astronomy (both 4.4) have the highest scores. Respondents from all of the disciplines investigated are positive about the contribution that AI makes to collaborating across the boundaries of disciplines. ‘I certainly see opportunities for AI in the humanities’, responded a historian to one of the open questions. ‘Especially in the more applied disciplines such as archaeology and linguistics. Nevertheless, I place major question marks against the value, opportunities, utility and ethics of AI in my discipline. The questions that people pose to AI are only as good as the people posing them.’ The quality of those questions could be improved if the Dutch system for research funding and facilitating creativity and out-of-the-box thinking were to be valued more, added a researcher.

Ai en de wetenschap. Diagram3.PNG
Ai en de wetenschap. Diagram4.PNG

AI is not yet a settled technology

It was not just the value and utility of AI for their own discipline that led to question marks among some respondents; the actual survey elicited quite a lot of responses too. Phil Macnaghten, Professor of Technology and International Development at Wageningen University, stated that AI is all too easily presented as if it were a settled technology. ‘But the reality is that AI is still al long way from achieving that. Just like other new technologies, it elicits a large number of social and ethical questions that have not been answered yet. In my opinion, a survey like this overlooks that reality.’

‘I’m concerned about how the term AI is used as a hype, as if it could solve all of our problems’, responded Christof Francke. He is associate professor of Bioinformatics at the HAN University of Applied Sciences and emphasises that AI is not a self-serving technology. ‘There is a strong need for people who have a clear knowledge of the disciplinary context and know which IT or data technique could possibly lead to solutions. In brief, there is a big demand for investments and solutions to real problems.’

Misleading and unscientific

Francke is definitely not the only person to use the term hype, and in responses to the open questions, the word occurs with striking frequency. ‘I’m enthusiastic, but I also think that it’s a bit overdone,’ said a respondent from the discipline of medicine. ‘We’re not in the slightest bit critical about what it exactly means and relatively little has been properly validated.’ A confirmation of this comes from no less than the field of computer science: ‘The term AI is misleading and unscientific. For me, it’s nothing more than a marketing hype around basic applications of statistical machine learning.’ An astronomer stated that first of all, more hands-on experience is needed before we can claim with certainty that AI plays a leading role in research. A biologist also used the term “hype” but nevertheless thinks that it is here to stay. ‘The problem is that policymakers think that they understand the term AI. This means that money goes to AI at the expense of other types of research.’

Critical anthropological research into AI

In Marketeffect’s survey, there is a wide range of opinions about the allocation of possible extra funding. The proposition that the Netherlands should make extra investments in AI research, even if that is at the expense of funding from other disciplines, elicits a slightly more negative than positive response among researchers: 39 versus 37 percent. Historians (60 percent negative versus 14 percent positive), philosophers (59 against 16), mathematicians (58 against 26) and political scientists (46 against 22) reject the option on the whole. Respondents from the disciplines of computer science and medicine, however, agree more often with the proposition.

The tone is mainly one of technological optimism

‘AI’s entry is inevitable, and society will probably fundamentally change’, responded a behavioural scientist in the survey. ‘How will AI influence our human intelligence, emotions and cognition? I see many fundamental, philosophical and ethical questions, and I think that it is important the research funding is made available for these and similar questions. At present, the tone is mainly one of “technological optimism”. NWO should create a lot of opportunities for critical anthropological research into AI. The question should be: what opportunities does AI offer us?’

Ai en de wetenschap. Diagram5.PNG

Online survey

In January 2020, Markteffect, at the request of the editorial board of Research, conducted a survey about how researchers view developments in the area of artificial intelligence. Markteffect approached researchers from eleven different disciplines spread across the NWO domains, who submitted a funding proposal to NWO between 2016 and 2020. From over 14,000 surveys sent online, 1484 fully completed surveys were received: a response rate of 10.5 percent. The disciplines investigated: Public Administration/Political Science, Biology, Philosophy, Medicine, History, Computer Science, Law, Religious Studies and Theology, Astronomy/Astrophysics, Engineering Sciences and Mathematics.

Download the article (in Dutch)

Tussen hoop en hype.PNG