EU’s AI Act: Why do we need it?

Part 1: The age of artificial intelligence (AI)

 

The EU is currently in the process of debating a proposed law that is known generally as EU’s AI Act. This law is being debated in a political and intellectual context whose complexities are not always transparent to a large number of those affected. Often these complexities deal with notions and artefacts that are simultaneously familiar and opaque. The result is a slipperiness of issues that can cloud what is at stake in the sporadically contentious and voluble debates around artificial intelligence (AI) in general and the EU’s attempt to regulate the sector, through its AI Act, in particular.  Yet, depending on how far into the future one wants to see, this debate can be momentous in political, legal, ethical or even existential terms.

Context: the age of artificial intelligence

In 2021 the UK government declared ours to be the ‘the age of artificial intelligence‘ (OAI 2021). It issued a national strategy to make the country an AI superpower by 2031. In this the UK was only doing what others had done earlier. In 2017 China had issued a State Council Notice that boldly declared the country’s intention to seize the ‘first-mover advantage’ in the development of AI because, the Notice stated, ‘the rapid development of artificial intelligence (AI) will profoundly change human society and life and change the world’ (Webster et al 2017). By March 2021 the United States’ National Security Commission on Artificial Intelligence was concerned enough to state that ‘China is a competitor possessing the might, talent and ambition to challenge America’s technological leadership, military superiority, and its broader position in the world’ (NSCAI 2021). Thus, in the past half-decade the AI race amongst nations has begun to boil over.

In this climate, Vincent C. Müller points out there has been no dearth of ‘reports and white papers’ and also ‘good-will slogans (“trusted/responsible/humane/human-centred/good/beneficial AI”)’ (Müller 2023). However, the EU was the first political entity to propose to legislate on the subject in 2021 while other political powers, such as the U.S.A. and China, with overwhelmingly significant presence in the AI field, had chosen to issue regulations and guidelines but not legislation. Even though in August 2023 China once again moved ahead of the field to push through a law on Generative AI (Roberts and Hine 2023), the EU’s proposal retains a special significance for democratic societies everywhere. This is because the varying approaches amongst nations are underpinned by deep and far-reaching attitudes and assumptions about the role of technology in society today.

Against this background, this series of blogs uses the text and deliberations made available by the EU’s proposed AI Act as a starting point to explore the meaning and challenge of artificial intelligence to society. To consider the nature of the challenge, it is useful to begin with defining what is covered by the term artificial intelligence.

What is Artificial Intelligence (AI)?

The difficulties in defining AI can be gauged from the fact that between April 2021, when the initial proposal (Proposal from here on) for this legislation was drafted by the European Commission, and June 2023, when the European Council and the European Parliament got involved in a ‘trilogue’ to hammer out the Act’s final wording, three different definitions of AI emerged from the three bodies.

In April 2021, the Proposal had simply defined an AI system for the purposes of regulation as ‘software that is developed with one or more of the techniques and approaches’ (AI Act 2022, 16h53  199/681) [emphasis mine] that the Proposal had listed in Annex I of the draft legislation.

In June 2023, however, to capture the fast-paced changes that had occurred in the field, the European Parliament, which represents the people of the member states of the EU, widened the definition from software alone to say that ‘(AI system) means a machine-based system’ (AI Act 2022, 16h53  199/681) [emphasis mine]. In its turn, the European Council, which represents the member states, proposed a definition that expanded even further to say that an ‘artificial intelligence system (AI system) means a system’ (AI Act 2022, 16h53  199/681) [emphasis mine]. This reflects the difficulty of encompassing the very wide range of technologies that have evolved under the umbrella term ‘artificial intelligence’ which can be software alone, or software incorporated in machines like robots, or software incorporated in other larger systems that have both AI and non-AI parts. This is why all three bodies have ultimately chosen to stress that ‘Artificial intelligence is a fast evolving family of technologies‘ (AI Act 2022, 16h53  11/681) [emphasis mine].

In terms of use, AI systems are already all around us. In the majority of the Western world they can form part of the lifeworlds in administration of justice, autonomous transportation, dangerous jobs like mining, domestic spaces, education, gaming and virtual reality, healthcare, industrial manufacturing, military, policing, repairing and enhancing the human body, robotics, security, Smarthomes, space exploration, space travel, and most recently, and controversially, the creative industries (Anguiano and Beckett 2023).

This dizzying variety makes it arduous to create general rules and regulations that address the full implications of using AI in all of the situations where they are currently being deployed, and the new uses of the technology that are being developed on an equally regular basis. More pressingly, the risks and dangers posed by some AI technologies are under-appreciated or even, due to their resemblance with human behaviours, hard to control even when recognised.

Under the circumstances, classifying AI systems by the risk they pose is an emerging strategy for creating differentiated regulatory systems by political authorities. However, as a recent report pointed out, classifying AI on the basis of risks posed is less straightforward than might be assumed. The report asserted that for a full 40% of the AI systems, it was difficult to define whether they fell into the high-risk category or not (IAAI 2023). When experts are daunted by the task of classifying and regulating AI, it is only natural for the layperson to feel out of their depth.

At the same time, the speed and reach of these technologies is already such that the task of shaping a societal response to them cannot be left to some unspecified point in the future. If a wider public response to them is to be meaningful, rather than guided by elite discourse alone, there is a need to ensure that the ordinary citizen can understand these technologies in their context. That is the only guarantee that both policy and action from all actors involved will converge to promote wider human welfare. Ultimately,  ‘policy is not just an implementation of ethical theory, but subject to societal power structures … ‘ (Müller 2023). Without adequate public awareness of the issues and options, ‘[T]here is … a significant risk that regulation will remain toothless in the face of economical and political power’ (Müller 2023).

The historical development of artificial intelligence technologies

To fully appreciate the nature and extent of what is involved in dealing with this family of technologies, this may be an opportune moment to take a brief look at the incredible journey it has traversed in the past 70 years or so, which has helped to inveigle the AI family into the daily lives of a great number of human beings without the majority even being aware of it.

It was in the 1940s that scientists, who had until then concentrated on developing computers that could calculate faster and faster, began to report on research that could map the working of the human brain in mathematical terms. The phenomenal potential of this possibility excited neuroscientists and mathematicians alike. Some consider the crucial idea for the future development of artificial intelligence to have come from a paper titled ‘A logical calculus of the ideas imminent in nervous activity‘ by Warren S. McCulloch and Walter H. Pitts in 1943.

In this paper, McCulloch and Pitts put forward ideas that could help to solve the age-old ‘Mind-Brain problem’ (MBP) in philosophy and science both. The MBP revolves around the notion that the intangible processes that constitute the human mind (thoughts and consciousness) can be explained through the physical activities of the human brain (which is accessible to researchers and machines for replicable investigations). Although this paper was very far from providing the techniques such as ‘neural mapping’, ‘machine intelligence’ and ‘complex system dynamics’ that are part of AI researchers’ repertoire today, it sparked ideas that would wind their way through many iterations to culminate in the current state of AI (Issit 2020b).

From this point on, it was possible for researchers to look into reducing thoughts, opinions, judgements and views held by particular individuals to some series of simple mathematical propositions such as ‘A equals B’. These kinds of simplified data could be handled by machines that essentially worked by making simple calculations at astronomical speeds with more and more data connecting the data points into complex patterns that defied clear unravelling by humans due to the sheer speed at which the data points were connected.

To give an extremely simplistic hypothetical example:

• the data point ‘person A has the same preferences as person B’ [from the dataset ‘similarities and differences amongst the employees of Company X’] connects to
• data point ‘B prefers toast to eggs’ [from the dataset ‘social media chat about food preferences in the UK’] connects to
• data point ‘toast is made of wheat’ [from the dataset ‘grain components of main food types consumed in the UK’] connects to
• data point ‘the Ukraine war has pushed the price of wheat up’ [ from the dataset ‘the impact of the Ukraine war’] connects to
• data point ‘food prices in the UK have gone up’ [from the dataset ‘the impact of the Ukraine war in the UK’]

• which can provide A’s boss with the insight that A’s recent loss of energy may be due to increased food bills that she has to manage.

In 1943, however, this kind of scenario was perhaps not even imagined, though the seeds for its possibility were sown by McCulloch and Pitts by suggesting that the ‘mind can be reproduced computationally’.

Another one of the varied routes that would lead to the present-day family of AI was also pioneered in the 1940s. In 1949 William Grey Walter produced a mechanical ‘tortoise’ that had been modelled to mimic complex behaviour not in humans but in animals. Rather grandiosely, these automated animal robots were called Machina speculatrix or, in a more fun way, simply ‘tortoise’. Walter’s article about his inventions was presciently titled ‘An imitation of life‘. The robots, at a very primitive level, did just that: they imitated behaviour of sentient beings, without having any sensory data themselves. They could indicate ‘hunger’, or ‘cry’ or ‘dance’ merely in response to electrical/mechanical/mathematical stimuli. They were machines that could seem alive. Walter may well have been unaware that this illusion of sorts would have implications that the EU’s AI act is trying to grapple with three-quarters of a century later.

Moving forward, the following two decades saw researchers take these insights into new directions. Herbert Alexander Simon, Allen Newell and John Clifford Shaw achieved the goal of making a computer ‘think’. Using a program called ‘Logical Theorist’ and employing the concept of ‘heuristic problem solving’, these researchers showed that a computer programmed in a certain way could make decisions that allowed it to prove 38 of the first 52 theorems in chapter 2 of the mathematical tome Principia Mathematica. This computer was not just imitating sentient behaviour; it was applying given principles of logic to solve complex problems. It was still very far from the thinking capacities of a human being but the question to ask was: were these machines imitating not just human behaviour but also human thinking? If so, was the hitherto somewhat mystical substance of a person’s ‘consciousness’ – intangible thoughts and reasoning – now grounded in a knowable, replicable algorithm that was open to human manipulation?

Yet, the computers lacked practical applications beyond research and academia and these questions were rarely asked by the general public. It was robotics with its promise or threat, depending on one’s point of view, that made more of an impression on the public’s consciousness. Nevertheless, the study of AI flourished in elite academic circles. In 1956 the first conference on AI took place. It was organised by pioneers Marvin Minsky, John McCarthy and Claude Shannon. The term ‘artificial intelligence’, credited to McCarthy, was also used at this conference. The institutionalisation of the discipline was reflected in the establishment of the first department for AI research at Massachusetts Institute of Technology (MIT) in 1959 (Issit 2020b).

‘Computing, machinery and intelligence’

In the UK, meanwhile, in a seminal article written in 1950, Alan Turing, often described as the father of modern computing, set out a detailed scenario of ‘digital computers’ that could be programmed to behave like ‘learning machines’ or ‘child machines’. At the heart of his vision was the idea of reducing the communication processes involved in teaching and learning amongst humans to a set of repeated codes or instructions such that the end result was not always the same fixed answer but rather the answer was always achieved through the same process.

Turing, however, went further in his imagination and envisioned the complexity produced by such a ‘learning’ programme to be used for communicative purposes that had hitherto been considered the most distinguishing trait or characteristic of humanity, setting it apart from all other animate and inanimate beings. An example game described by him in this article has since come to be known as the ‘Turing Test‘
that is still used for testing how successfully a computer can make a human believe that they are communicating with another human being. In this, Turing changed the idea of artificial intelligence from a ‘structural’ to a ‘performance-based’ concept; the test of such intelligence was how it was experienced by the recipient rather than what kind of hardware or software needed to be incorporated to replicate human mental processes exactly (Issit 2020a). Most crucially, however, Turing hoped ‘that machines will eventually compete with men in all purely intellectual fields’ (Turing 1950).

In his article, Turing also laid out two different ways in which thinking and learning machines that competed with ‘men’ intellectually could be built: the first was to equip ‘digital computers’ with reasoning power as described earlier. To do this, they needed to be taught like humans teach each other. Here Turing gave the example of ‘Miss Helen Keller‘ which to him showed that all that is needed for ‘education’ to ‘take place’ is that ‘communication in both directions between teacher and pupil can take place by some means or the other’. He specifically clarified that there was no need for ‘legs, eyes etc.’ for educating someone to be ‘supercritical’. Being ‘supercritical’, which clearly comes across as a laudable objective in the article, is reflected in having a mind which, when presented with an idea, can ‘give rise to a whole “theory”‘. Turing asked with no hesitation, ‘Can a machine be made to be supercritical?‘ (Turing 1950).

The second path briefly touched on in the article was ‘to provide the machine with the best sense organs that money can buy, and then teach it to understand and speak English’. This process of combining ‘sensory data’ with ‘machine parts’ may be developing somewhat differently in recent years with computer chips having been implanted in human brains in some recent experiments for therapeutic reasons. The point to note here is that this process allowed for the manipulation of human behaviour though, it must be stressed, this was at present was for therapeutic reasons only (Harari 2016, p. 334). The concerns raised by this development in some quarters, however, will be discussed in later blogs.

Turing’s worldview

Reading Turing’s article for those over 35 years old presents a sense of being strangely disorientated. This is because despite the article being about a relatively technical aspect of one of the sciences, something that was until recently taken to be removed from the heated hurly-burly of the socio-political world, the article contains elements of a worldview that have recently been recognised as problematic. Yet, significantly, this worldview was not Turing’s alone, but was widespread until recently, and still remains embedded in large swathes of the population around the globe.

At the beginning of the article itself (on p. 2), Turing poses the question whether the subject of the article ‘Can machines think?’ is worthy of investigation. To establish that it is, the article contains a detailed discussion of the various objections often made by different groups against such an activity.

These objections fall into 2 categories: the ‘morality’ category and the ‘possibility’ category. The first covers objections on ‘theological’ grounds that suggest that in creating machines that can think humans will be ‘usurping’ the divine power of conferring souls, which belongs to God alone. The objection, therefore, is that machines cannot think because it is against divine will; the implication is that humans should not attempt to make such machines. However, Turing is not convinced. His argument seems to be that the success of such an endeavour can be taken to flow from God’s will too.

Most of the other objections considered by him fall in the ‘possibility’ category, i.e., objections that claim on a variety of grounds that it is not possible for machines to think in the way that human beings can. Turing’s answer seems to be that whether machines think ‘like’ a human being or not is besides the point so long as they can follow the process involved in human thinking to produce results similar to it.

Turing acknowledges that he has ‘no very convincing arguments of a positive nature to support’ his views that it is both possible and desirable to make ‘thinking’ (or ‘learning’) machines (p. 17). However, his tone is decidedly sympathetic to the machines in question. A close reading of the article reflects a worldview that valorises an intellectual elite possessing ‘supercritical’ intellectual abilities. Thus the enterprise of making machines that can think is being advocated despite the understanding that such ‘supercritical’ machines, should they come into existence, will be able to intellectually ‘outstrip’ the majority of humans who Turing states baldly to be ‘subcritical’ (p. 17).

This is a worldview where the powers acquired through the practice of science for ‘improving’ the world are held in high esteem. By using the higher reasoning faculties of the human mind, scientists and engineers were stealing a march over nature. They could speed up the process of ‘evolution’ or the ‘survival of the fittest’ amongst their creations by expeditiously removing any weaknesses they detected in their machines rather than wait for the random process of selection that works much more slowly in natural evolution. It was this excitement of being able to make the world a better place for human habitation that gave science, and its adjunct technology, widespread respect.

In the second decade of the 21st century, the respect for science and technology are acquiring a more complex and nuanced character amongst a growing number of people. The beliefs about the human mind that are reflected in a simple division of human beings into ‘supercritical’ and ‘subcritical’ are also being supplanted by understandings of neurodiversity or the different ways in which human brains can be wired, giving them different capabilities.

However, the feeling of disorientation in reading this article in 21st century stems from the fact that Turing’s excitement about machines was a shared worldview that is perhaps still widespread. His admiration in considering the possibility of creating an entire human being from a single human cell as ‘a feat of biological technique deserving of the very highest praise’ is still likely to echo around the world without too much opposition.

The grand narrative of science and technology as civilization’s highest achievement continues to be part of a way of seeing reality that is largely undisputed. As recently as the turn of the century, modernity as reflected in the cultural states obtaining in Western Europe and the United States, was taken to be the teleological goal for the fullest development of all societies everywhere, by the majority of human beings on the planet.

The backbone of these cultural states was a high level of technological advancement growing out of a rational use of scientific knowledge. The pursuit of what was taken to be ‘dispassionate’ scientific knowledge was considered a noble calling with the typical scientist seen as a disinterested practitioner dispensing universally established ‘truths about the natural world’ (Sismondo 2010, p. 18). This view of science and technology underpins Turing’s article and its advocacy of striving to build ‘supercritical’ machines that will compete with ‘man’ [sic] in all intellectual endeavours’. Even though this view is still deeply ingrained in a large number of people around the world, it is doubtful how many of them are attentive to its full implications for the majority of human beings (Turing himself devoted only one sentence in a 21 page article to mentioning ‘subcritical’ humans and none whatsoever on what ‘supercritical’ machines competing with such humans will mean for them).

It is, therefore, with an act of sheer will that people have to engage with the warnings that are beginning to filter through about the threat and dangers posed by such ‘thinking’ machines, or artificial intelligence, now that they have actually appeared on the scene, even if they are not yet capable of going from an idea to a whole theory by themselves. Despite the nation’s long history of being at the forefront of a rational, scientific modernity, the UK’s Prime Minister Rishi Sunak was impelled to warn in November 2023 that though he did not wish to be ‘alarmist’, he felt that [M]mitigating the risk of human extinction from AI should be a ‘global priority’  (Gregory and Kleinman 2023).

Humans and technology

The discomfiture about this kind of technology, despite its highly rational nature and astounding capabilities, is clear in various passages of EU’s proposed AI Act. The draft Act reiterates in many places that AI systems must be ‘human-centric’. It prohibits AI deploying ‘subliminal techniques’ that can adversely affect the power of autonomous choice in humans ‘in ways that people are not consciously aware of, or even if aware not able to control or resist’ (Recital 16, EU Council version). Further, ‘Article 5 prohibits systems using subliminal techniques that modify people’s decisions or actions in ways likely to cause significant harm’ (Bermúdez et al 2023).  The need to clearly spell out the centrality of ‘human’ interest in this new world and even the urgency to protect the capacity of humans for autonomous choice in the face of possible control by ‘rational’ but ‘not-human’ agents is a far cry from Turing’s optimistic and accepting discussion of machines that are our intellectual competitors.

The EU is not alone in its alarm at the unconventional threats to humans from AI technologies. In the United States, the NSCAI accepted that the use of AI technologies will change the ‘dynamics within human-machine “teams”’ because until now machines performed within ‘a clearly defined set of parameters or rules programmed by a human’. However, as AI technologies evolve

‘computers will be able to learn and perform tasks based on parameters that humans do not explicitly program, creating choices and taking actions at a volume and speed never before possible.’

The report recognises explicitly that the ‘preservation of individual liberties’ is in danger by calling for ‘continued vigilance’ to safeguard it.

Amongst the main AI leaders, China too, has stated in its draft of Artificial Intelligence Law, Model Law v. 1 that ‘AI should be people-centered and direct intelligence for good. Ensure that humans can continuously supervise and control AI, with the ultimate goal of always promoting the welfare of humanity.’ This is in contrast to China’s earlier response to AI technologies in its ‘New Generation Artificial Intelligence Development Plan’ of 2017 where the focus was very much on harnessing AI technologies for the country’s socio-economic development.

The change in China’s stance to come closer to the EU’s in stressing the primacy of ‘human’ interests in deploying AI technologies is perhaps a measure of how in the past year the world has raced away from the worldview embodied in Turing’s brilliant article. It seems clear that the question of the relationship between humans and ‘thinking machines’ has taken on a new urgency.

It is a question, however, that has been pondered on by a small number of thinkers and practitioners for more than a century even as science-centric worldviews, like those reflected in Turing’s article, have dominated the cultural space. The next blog in the series looks at some of these viewpoints to ask whether they have the answer to our present dilemmas of how to compete with machines competing with us.

 

 

 

References

Anguiano, D. and Beckett, L. ‘How Hollywood writers triumphed over AI – and why it matters,’ The Guardian, 1 October 2023. Available at: https://www.theguardian.com/culture/2023/oct/01/hollywood-writers-strike-artificial-intelligence (Accessed: 15 November 2023).

‘AI Mandates (20 June 2023)’ : Proposal for a REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL LAYING DOWN HARMONISED RULES ON ARTIFICIAL INTELLIGENCE (ARTIFICIAL INTELLIGENCE ACT) AND AMENDING CERTAIN UNION LEGISLATIVE ACTS 2021/0106(COD) 20-06-2023 at 16h53. Documents. [online] Available at: https://artificialintelligenceact.eu/documents/ (Accessed:10 November 2023).

Bermúdez, J. P., Calvo, R. A., Nyrup, R. and Deterding, S. ‘The AI Act needs a practical definition of “subliminal techniques”‘ Euractive. com, 1 September 2023. Available at: https://www.euractiv.com/section/artificial-intelligence/opinion/the-ai-act-needs-a-practical-definition-of-subliminal-techniques/ (Accessed: 10 November 2023).

European Commission (EC) (2021) Proposal for a Regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) and amending certain union legislative acts {SEC(2021) 167 final} – {SWD(2021) 84 final} – {SWD(2021) 85 final}. Available at: https://artificialintelligenceact.eu/the-act/ (Accessed: 16 November 2023).

Frankenfield, J. ‘The Turing test: What is it, what can pass it, and limitations,’ Investopedia, 31 July 2023. [online] Available at: https://www.investopedia.com/terms/t/turing-test.asp#:~:text=A%20Turing%20test%20works%20with,give%20the%20response%20or%20not (Accessed: 16 November 2023).

Gregory, J. and Kleinman, Z. ‘Rishi Sunak says AI has threats and risks – but outlines its potential,’ BBC News, 26 October 2023. [online] Available at: https://www.bbc.co.uk/news/uk-67225158 (Accessed: 30 October 2023).

Harari, Y. N. (2017) Homo Deus. Kindle edn. Vintage.

Issit, M.L. (2020a) ‘Automated ethics: Robotics, ethics, and morality (1920s–1950s),’ in Opinions Throughout History: Robotics & Artificial Intelligence, 1st edn. Grey House Publishing. Available at: https://search.credoreference.com/articles/Qm9va0FydGljbGU6NTYxMzg=?aid=275203 (Accessed: 16 November 2023).

Issit, M.L. (2020b) ‘Approaching intelligence: First steps toward modern artificial intelligence (1940s–1960s)’, in Opinions Throughout History: Robotics & Artificial Intelligence. 1st edn. Grey House Publishing. Available at: https://search.credoreference.com/articles/Qm9va0FydGljbGU6NTYxMTY=?aid=275203 (Accessed: 29 October 2023).

Issit, M.L. (2020c) ‘Conclusion: Robotics and AI: Oppressors or a new utopia?,’ in Opinions Throughout History: Robotics & Artificial Intelligence. 1st edn. Grey House Publishing. Available at: https://search.credoreference.com/articles/Qm9va0FydGljbGU6NTYxMzg=?aid=275203 (Accessed: 11 November 2023).

Initiative for Applied Artificial Intelligence (IAAI) (2023) ‘AI Act: Risk classification of AI systems from a practical perspective’. [online] Available at: https://www.appliedai.de/en/hub-en/ai-act-risk-classification-of-ai-systems-from-a-practical-perspective (Accessed: 10 November 2023).

McCulloch, W.S., Pitts, W. A. (1943) ‘Logical calculus of the ideas immanent in nervous activity,’ Bulletin of Mathematical Biophysics vol. 5, pp. 115–133. Available at: https://doi.org/10.1007/BF02478259 (Accessed: 15 November 2023).

Müller, Vincent C. (2023) ‘Ethics of artificial intelligence and robotics,’ The Stanford Encyclopedia of Philosophy, Fall 2023 edn in Edward N. Zalta and Uri Nodelman (eds.), https://plato.stanford.edu/archives/fall2023/entries/ethics-ai/ (Accessed: 30 October 2023).

National Security Commission on Artificial Intelligence (NSCAI) (2021) ‘Final report’. Available at: https://reports.nscai.gov/final-report/introduction (Accessed: 29 October 2023).

Ng, K. W., Zhou, J., Murphy, B., Creemers, R. and Dorwart, H. (ed.) ‘Artificial intelligence law model law v. 1.0 (expert suggestion draft)’. Available at: https://digichina.stanford.edu/work/translation-artificial-intelligence-law-model-law-v-1-0-expert-suggestion-draft-aug-2023/ (Accessed: 24 December 2023)

Office for Artificial Intelligence (OAI) (2021) ‘National AI strategy’. Available at: https://www.gov.uk/government/publications/national-ai-strategy/national-ai-strategy-html-version (Accessed: 15 November 2023).

Roberts, H. and Hine, E. (2023) ‘The future of AI policy in China’, East Asia Forum, 27 September. Available at: https://www.eastasiaforum.org/2023/09/27/the-future-of-ai-policy-in-china/#:~:text=On%2015%20August%202023%2C%20a,that%20specifically%20targets%20generative%20AI (Accessed: 28 December 2023).

Turing, A. M. (1950) ‘Computing machinery and intelligence,’ Mind, vol. 49, pp. 433–460. Available at: Article 1 in Google search here (Accessed: 30 October 2023).

Walter, W. G. (1950) ‘An imitation of life,’ Scientific American Magazine, vol. 182, no. 5 (May 1950), p. 42. Available at: https://www.scientificamerican.com/article/an-imitation-of-life/.

Webster, G., Creemers, R., Kania, E. and Triolo, P. (trans.) (2017) ‘China’s “New generation artificial intelligence development plan (2017)”’. Available at: https://digichina.stanford.edu/work/full-translation-chinas-new-generation-artificial-intelligence-development-plan-2017/ (Accessed: 15 November 2023).

 

 

 

 

 

 

 

 

 

 

Do data always have a ‘correct’ interpretation? Reading a clash in data for conceptions of culture, authority and alternative worldviews

I recently had the opportunity to engage with a group of academics puzzling over the meaning of a large dataset that did not make straight-forward sense. The group who had collected the data were experienced academics from a Western university. The data had come from a very large group of teachers in a rural setting in South Asia. These teachers had been invited to digitally train on a new method of teaching some basics of their subject. The training had been presented as a Professional Development Activity and carried a certificate of completion.

The Western academics had been careful to deliver the training and collect follow-up data through competent local agents. To assess the translation of the training into daily teaching, a survey of 1000 participants was carried out soon after the training had been delivered. A large majority of the participants answered in the affirmative on a questionnaire asking if they were incorporating their new knowledge into daily practice. The results were not surprising because no special equipment or digital facility was required to include the new method in classroom practice.

However, in an effort to drill down into the data and understand what had worked, the academics also carried out detailed interviews of about one-fifth of the survey participants. They were taken aback to find that the results of this qualitative data-collection exercise returned negative results. They suggested that, contrary to the survey results, the majority of the participants were not using the new training. It was this contradiction and how to address and interpret it that the academics were puzzling over at the session I happened to attend.

The discussion was lively and interesting; some other academics, who had not been part of the data-collecting group, also confirmed their own previous experience of similar contradictions when undertaking research in the region. It was easy to see how the contradiction in the data carried a whiff of dishonesty on the part of the participants though the data-collecting group were wary of jumping to conclusions. Still, some of the researchers present voiced their opinion clearly that only one result could be ‘correct’ and they felt that the interviews were much more likely to have shown a true picture and should be given greater credence. Why, then, did the participants answer in the affirmative on the survey questionnaire?

Making data speak

Amongst a minority of those present, including me, there was a sense that data do not always end up answering the question asked of them, or at least, not only what is asked of them. Sometimes more insightful data interpretation requires changing the lens through which a dataset is viewed. One such scholar suggested that a local notion of ‘politeness’ might have played a role in the large affirmative survey result. The researchers were also kind enough to engage with my observation that perhaps neither one of the datasets was ‘wrong’: thinking of my own Indian background, I was curious to know whether the researchers had thought about the larger cultural context of their research. Had they, for example, looked into what ‘Professional Development Activity’ meant to the teachers being trained? In answer, one of the academics pointed out that the pupils these teachers were teaching had been found to struggle with some very basic learning on a number of occasions so the need for professional development should have been self-evident. Another academic confirmed that their group had been confused by enthusiastically-organised professional development sessions on earlier occasions too where, for some strange reason, very little time had been allocated for actual professional development activities. They persevered, however, believing that, the very holding of the sessions suggested that a space existed where, with time and effort, professional development could be supported. Yet these contradictions in the attitudes of the local teachers made them pause and wonder about the nature of this space and the best way of using it to meet teacher-development goals.

At this point I would like to quote the post-colonial scholar Gayatri Chakravorty Spivak to frame the part-reflection, part-analysis of this question that follows. In her ground-breaking essay ‘Can the subaltern speak?’, Spivak had discussed general theory with examples drawn from Indian history because, in her words: ‘I turn to Indian material because, in the absence of advanced disciplinary training, that accident of birth and education has provided me with a sense of the historical canvas, a hold on some of the pertinent languages that are useful tools for a bricoleur…’ (p. 281).

‘White men saving brown women from brown men’

Staying with Spivak’s essay, I want to draw attention to an  example of ‘differend’ – the ‘inaccessibility of, or untranslatability from, one mode of discourse in a dispute to another’ (p. 301) used by her. Contesting Foucault and Deleuze’s position, which seemed to suggest that if the intellectuals made way for the ‘subaltern’ to speak directly, then a ‘pure’ subaltern consciousness, free of ideology, would express itself. To clarify, in our case, the term ‘subaltern’ does not apply; I do, however, think of the ‘data-providers’ as the less powerful side in academic data-collection exercises. In this, I take knowledge and control of the process to be sources of power that reside with the designers of the exercise, who also collect data on parameters selected by them.

Spivak discusses the difficulty of interpreting the abolition of sati by the British rulers of India in 1829. Sati was a practice that was neither widespread nor did it have any particular class- or caste-based significance attached to it. It was an extreme practice, though, in which a Hindu woman chose to self-immolate on her dead husband’s pyre in an act of marital devotion and ‘purity’, rather than live on without him.

Even though sati was not widely practiced, it carried a high emotional charge and the women committing sati were venerated as martyrs who had displayed the highest example of feminine virtue. The abolition of sati was carried out despite nativist opposition and was justified by the British in moral terms as securing the helpless woman’s freedom to live on despite her husband’s death. In Spivak’s frank analysis it ‘has been generally understood as a case of “White men saving brown women from brown men”‘ (p. 298).

Against this, however, Spivak also cites the nativist Indian argument that insisted in equally moral terms that the ‘women actually wanted to die’ (p. 298). To the advocates of sati, the abolition denied a native woman the freedom to conduct herself in accordance with her religious and moral beliefs and was thus a violation of a moral law, seen differently. It is here that Spivak cites Jean-Francois Lyotard’s term ‘differend’  (p. 301) as perhaps the right one to understand these clashing viewpoints on women’s ‘sacrifice’ of themselves.

What stands out in Spivak’s analysis is that the woman’s voice is not available; others speak for her. Further, her point is not to question the abolition of sati as a good thing but rather to point out the difficulties of reading an act so clear and, one would assume, transparent as a woman burning herself on the pyre of her dead husband.

Again, the point that comes through is that even if the women had been given a chance to speak, they would have spoken with the ideas available to them. Both the British and the native Indians spoke out of their own contemporary traditions. The Indian position has been shown as ‘superstitious’ and ‘patriarchal’ through more than a century of history-writing.

Interestingly, Spivak makes clear the ideological nature of the British position too. She writes: ‘Perhaps sati should have been read with martyrdom, with the defunct husband standing in for the transcendental One; or with war, with the husband standing in for sovereign or state, for whose sake an intoxicating ideology of self-sacrifice can be mobilized (p.302).’ Dying for one’s ‘nation’ was more modern and acceptable to the British but not necessarily any more rational than a woman sacrificing herself for a religious ideal.

In the context of our current discussion, cultural influences and the difficulty of interpreting them can be equally pertinent to less dramatic acts of answering surveys and questionnaires.

What complicates cross-cultural interpretations in South Asia even further is the situation that Homi Bhabha, and other scholars, have tried to capture in notions of cultural ‘hybridity’ or ‘creolization’. The latter term is applied mostly to discussions about territories in the ‘New World’, however, Robert Young’s explanation of the process as an ‘unconscious hybridity, whose pregnancy gives birth to new forms of amalgamation rather than contestation’ (Ashcroft et al, 2000, p. 52) can be seen in South Asia too.

Bhabha asserts that these phenomena often occur in ”border situations and thresholds … where identities are performed and contested”. Instead of stressing the binary oppositions between groups and nations on national, cultural, racial or gender-based grounds, he is interested in how identities are negotiated, shaped and re-shaped at the faultlines of these differences. With nearly 200 years of imperial British rule in the region, such border situations were not hard to find.

Next, I would like to briefly look at Bernard S. Cohn’s exploration of the representation of authority in British India. This essay not only illuminates some of the mechanics of the processes involved in shaping hybridity or creolization, but it also touches on a concept relevant to analysing the conflicting responses from the teachers in the sample we were discussing, i.e., response to how authority is represented.

Cultural construction in interstitial spaces

Bhabha introduced the concept of ‘interstitial spaces’  to studying culture  to describe places and times where ‘the overlap and displacement of the domains of difference’ occur and where ‘intersubjective and collective experiences of nationness, community interest, or cultural value are negotiated’ (Bhabha, 1994, p.2).

Many cultural structures and values in South Asia have evolved in the contemporary period through a close interaction between the ‘traditional’ cultural norms and the form of Western modernity practiced by the imperial British rule in India from about 1757 to 1947. The comparatively small number of British who administered the authority of the Raj over a very large number of Indians meant that merely imposing new norms and values through force, as in the abolition of sati, was not as frequent an experience as often imagined, except for maintenance of British rule itself. Without denying the cruelties of subjection, it may be asserted that in the cultural sphere, contestation and negotiation were resorted to more often than might be supposed from an overview of the historical record.

In Cohn’s telling of the great Delhi Durbar, or imperial assemblage, of 1877 it is clear that it was an occasion when the British authorities took time and trouble to adopt the rituals of the Indian imperial courts to represent their authority to the locals in forms acceptable to the latter.

‘The princes of India bless you’

Discussing the ways in which the British rulers created a ‘ritual idiom’ for their authority in India, Cohn unpacks the meaning of the aforementioned Durbar, when Queen Victoria was proclaimed the Empress of India. By this Durbar the imperial authority of the defeated and deposed Mughal emperor was ritually replaced by the authority of the British monarch.

Those in charge of the Durbar were divided over the most effective way to proceed. The British had to constantly map an unfamiliar ‘sociology’ onto the cultural landscapes they were familiar with, i.e., society in Britain. Some of those involved viewed India as a feudal society with a ‘natural aristocracy’ of princes and local rulers worth cultivating to secure the obedience of the people through them.

The Viceroy, Lord Lytton, and his supporters therefore attempted to consciously create a new ‘idiom’ for the new rule. They deliberately hybridised British and Indian courtly forms and meanings. This was  to ensure that both parties to the ‘contract’ – the Indians and the British – could read the ceremonies by their own cultural understandings of  what the relationship between the sovereign and the people should be.

The need for such efforts was clear to Lytton and others.  Cohn gives several examples to show how, prior to the deposition of the Mughal emperor, a European mercantile eye had often misunderstood the deeply symbolic nature of their interactions with the ruling elite of India. One would suffice here: the main ritual at the Mughal court used to be ‘an act of incorporation’. At this, the persons being honoured by the ruler presented gold coins or other jewels and valuables to him as a peshkash, and he, in turn, presented them with similar valuables and clothing called khelat. The gestures carried symbolic meaning whereby the favoured person was ‘incorporated’ into the being of the emperor.

Outwardly, the ritual was carried out as an exchange of goods and valuables. Yet, there were other rules around the gifts exchanged (for example, the emperor’s gift was carefully preserved and displayed and not sold or given away by the recipients and their descendants) that raised it in the eyes of the Indians above a mere monetary exchange. The British, however, saw it ‘following their own cultural codes, and assumed there was a direct quid pro quo involved’ (p. 170).

In the period prior to the formal deposition of the Mughals, when the British  held nominal authority over the emperor’s territories, they kept the ceremony but were not averse to ‘recycling’ the presents they received. They gave to one Indian party what they had received from another, and sometimes sold the gifts by auction (p. 172). To quote Cohn:  ‘Mughal ritual might seem to have been retained but the meanings had been changed’ (p. 172).

Such cultural misunderstandings on a very large scale added fat to the fire that finally burned as  the First War of Indian Independence in 1857 (described as the Indian Mutiny by many British historians). British authority was re-established through much blood and tears on both sides in 1858. At the Durbar in 1877 the Viceroy was keen to assuage local feelings while  impressing the queen’s authority upon the people.

The  result was a form of  rituals at the Delhi Durbar that many experienced British civil servants found ‘”trivial and silly”‘ (p. 191). The official painter was horrified at the ‘garish’ mix-and-match in the scenario he was commanded to memorialise for posterity. Following the Mughal precedent, the attending Indian princes received presents from the imperial ruler, only the presents  did not take the form of khelat. Instead the princes received ‘shield-shaped banners’ and crests in ‘the European mode’ (p. 191) so that the Indian princes recognised the familiar rituals of imperial authority, as well as the changing nature of it.

Despite the horror of the detractors, the efforts seem not to have been in vain. Their success was evident at the Durbar itself when Sindhia, one of the most powerful princes of India, stood up and, it is believed, spontaneously addressed the queen as:

‘Shah in Shah, Padshah, May God bless you. The princes of India bless you and pray that your hukumat [rule]… may remain steadfast forever’ (p. 206).

The prince completely omitted addressing the queen by the new and unfamiliar title of kaiser-i-hind that had been decided on. However, what is of import here is his spontaneous gesture of loyalty. The viceroy’s efforts with the hybridised rites and rituals had made an emotional connection through recognised practice, enough to bring forth an emotional response. The lack of such emotional recognition had often set Indian princes on a path against the British which changed significantly after this ceremony. Cohn asserts that ‘the codification of the ritual idiom’ (p. 208) at the Durbar ‘set the terms of discourse of [even] the nationalist movement [that ended the empire] in its beginning phases’ (p. 209). Moreover, ‘[T]he British idiom did not die out easily or quickly,’ Cohn says, ‘and it may still be alive in various forms’ (p. 209).

Performing modernity

Agreeing with Cohn, I suggest that many South Asians in professional development contexts still draw upon a ‘cultural symbolic constitution’ that was forged in the interstitial spaces of the colonial era. A ‘cultural symbolic constitution’ shapes our engagement with ‘cosmologies, world views, ethical systems, legal codes, definitions of governmental units and social groups, ideologies, religious doctrines, myths, rituals, procedures, and rules of etiquette’ (Imden, quoted in Cohn 1983, p. 173). I suggest  that the conflicting data from the teachers that we were discussing can be explained in terms of the attitudes towards ‘authority’ that are part of the South Asian cultural symbolic constitution. In this constitution, in the way described by Cohn, sometimes the form of the representation of authority and the response to it is read as being more important than its substance.

First, even after 70 years of independence, in many places, particularly in small towns and rural areas, the respect for British, and by extension Western, knowledge retains its authority, even though it is often contested vocally. Thus, projects run by Western researchers, even through local agents, often carry this weight of authority.

Second, in cases where Western researchers approach participants through educational bodies, a further layer of authority is added.

Third, digital media carries its own form of authority as a symbol of that which is modern and powerful.

The Country Comparison tool of the Cultural Factor Group, building on the research insights of Professor Geert Hofstede, awards most South Asian countries a score of between 55 and 80 on the ‘Power-Distance‘ dimension. These high scores mean that the people’s attitude towards authority is: ‘dependent on the boss or the power holder for direction’. The pattern is of having a ‘paternalistic leader, [where] management directs, gives reason / meaning to ones [sic] work life and rewards in exchange for loyalty from employees’ [emphasis mine].

Under the circumstances, with three layers of authority tacitly driving the uptake of a course, it would not be surprising that a good number of participants took the expectation to do the course as the direction to do so.

However, the relationship of individuals with professional authority, and what it can expect from the individual, is a complex one in the region. As shown in the efforts made by the imperial British at the Delhi Durbar to connect to the ‘cultural symbolic constitution’ of the people, representations of authority carry distinct meanings to the recipients. No authority is likely to gain the support it desires if the forms of its representation are not fully recognised.

In our example, to the Western academics, it was seen as self-evident that teaching is a ‘profession’ in which professional development training should lead to better professional results, i.e., the improvement of learning by the pupils. However, in the regional subconscious, ‘the professional’ and the ‘teacher-pupil relationship’ belong to two different emotional spheres. They fall within two different classifications in the cultural-symbolic constitution that Imden says ’embraces such things as classificatory schema, assumptions about how things are’ (as quoted in Cohn, 1983, p. 173).

The idea of the ‘professional’ is governed by schemas about relationships between adults encompassing the world of work, livelihood and worldly power. The farther one moves from the large urban centres, the more the idea of the ‘modern’, which the ‘professional’ is part of, takes on a performative tone. What this means is that the modern, and hence also the professional, is not ‘“a stable identity or locus of agency from which various acts follow; rather, [it… is … instituted … through a stylized repetition of [habitual] acts” (Butler 1999, 179 as cited Mari, 2023). This would include meeting expectations about updating one’s professional knowledge and rather literally a mechanical ticking of the affirmative box on the survey questionnaire.

These acts of the ‘professional’ sphere carry respect so that many teachers would have been motivated to complete the course. Theoretical knowledge, degrees and formal awards are held in high esteem throughout the region. I cannot doubt that a majority of those teachers would have made sincere efforts to gain the knowledge, or sufficiently so, in that course, to pass the test for the certificate. However, that effort is born of their self-image as modern educated professionals who are in touch with the latest developments in the world of education as provided by prestigious Western education, and of their respect for authority. It does not always connect mentally to activities in the classroom.

The teacher–pupil relationship within the classroom is grounded in a different psychological classification. It leans into ideas about adult–child relationships that have a more moral and emotional character. The teacher stands in a relationship of grave moral and intellectual authority vis-a-vis their pupils. Many South Asian languages contain expressions similar in meaning to the popular Sanskrit mantra comparing the ‘guru’ or the teacher to the three supreme deities of Hinduism – Bramha (the creator), Vishnu (the generator) and Mahesh (the destroyer). Education, when framed in traditional terms, carries the same ‘aura of deep moral seriousness’ that Clifford Geertz ascribed to all things religious in his discussion of human cultures (Geertz, 1973, p. 136). This notion is more responsive to a language of morality, care and nurturance of a young life than professional development. Even though the expression of this care may come across as a harsh disciplinarian dismissiveness to the onlooker, it can be more easily engaged by the semiotics of the ‘moral’ than of the ‘professional’.

As for teaching, the teacher’s job is to pass on the knowledge of the time-honoured ‘authorities’ in a given subject, in time-honoured ways. It is for the pupils to show respect to the authority of the teacher and of the knowledge by learning. If a lack in learning occurs then it is a lack in the pupil and not the teacher.

This is not to deny that there may have been other reasons too for the non-implementation of the training in the classroom. My Point here is simply to offer an alternative explanation (rather than mass dishonesty which may suggest itself more readily) for the contradiction between the large number of affirmative answers on the survey asking whether the new training was being used in the classroom and the large proportion of negative ones in the more detailed interviews.

It is also not my intention to say that none of the Western effort is ever adopted the way it is intended by the Western providers. No. In any sample, there would be a proportion of teachers who would apply the course exactly as intended but their motivation may come from Western theory or from indigenous child-friendly ethos of paternalism more familiar to them from their own cultural mores.

The data we were discussing may be suggesting a need for building new professional development practices by connecting them more explicitly to local worldviews. Preliminary conversational sessions exploring local ideas about the role of education and of the teacher may be useful too. ‘Co-creation’ of professional development frameworks in place of the dissemination and simple sharing of knowledge developed in the West is likely to be more effective.

In 20th century India, Gandhi was able to mobilise underlying notions of sacrifice and simple living in the indigenous culture. It gave him the means to turn a national movement that until then had been confined to the educated classes into a mass movement of the people. As politicians are apt to remind us repeatedly, any campaign for change is a campaign for the hearts as well as the minds of the recipients.

 

References

Ashcroft, B., Griffiths, G. and Tiffin, H. (2000) Post-colonial studies: the key concepts, 2nd edn. London: Routledge.

Bhabha, H. K. (1994) ‘Introduction’ in The location of culture. London: Routledge.

Cohn, B. S. (1983) ‘Representing authority in British India’ in E. Hobsbawm and T. Ranger (eds.), The invention of tradition. Cambridge: Cambridge University Press, pp. 165–210.

Geertz, C. (1973) The interpretation of cultures, Kindle edition. New York, Basic Books.

Imden, R. (1976) ‘Cultural symbolic constitutions in ancient India’, mimeograph, pp. 6–8, as quoted in Cohn, B. S. (1983) ‘Representing authority in British India’ in E. Hobsbawm and T. Ranger (eds.), The invention of tradition. Cambridge: Cambridge University Press, p. 173.

Mikkola, M. (2023) ‘Feminist perspectives on sex and gender’, The Stanford Encyclopedia of Philosophy, Fall 2023 Edition in Edward N. Zalta and Uri Nodelman (eds.), URL = <https://plato.stanford.edu/archives/fall2023/entries/feminism-gender/>.

Perloff, M. (1998) ‘Cultural liminality/aesthetic closure?: The “interstitial perspective” of Homi Bhabha’. [online] Available at: The “Interstitial Perspective” of Homi Bhabha (upenn.edu). (Accessed 10 December 2023)

Spivak, G.C., ‘Can the subaltern speak?’ in C. Nelson and L. Grossberg (eds.), Marxism and the interpretation of culture. Basingstoke: Macmillan Education, 1988, pp. 271–313.

Photo by Alina Grubnyak on Unsplash