Decision-making skills in the fourth industrial revolution

Steering the buildup of the 4IR

The first kind of decision concerns the confluence of the technologies themselves. The basic question is whether this is a development to be desired as such, and if so whether technical and substantial preconditions should be drawn up.

The confluence of the newly developed technologies is only possible if huge amounts of data can be stored and transmitted at high speed. Currently, three developments are crucial: the investments in fiber optic cables, the launch of mini satellites to establish a global broadband network, and the building of hyperscale data centers (HDCs).

Fiber optic cables are already spread over the world since 1988 and keep on expanding to a global Fiber Optic Cable market size projected to reach 22 billion US $ by 2026 from 15 billion US $ in 2020. In 2018, more than 500 million fiber km were installed, and in 2019 another 480 million km. ¹

As to the satellite internet network, SpaceX started to launch 60 mini-satellites in 2019. At the end of 2021, it has already over 1800 satellites in Low Earth Orbit. It plans to launch 12,000 more in the short term and over 42,000 later, resulting in a grid that covers the globe and aims to enable an average internet speed all over the world of 134.7 Mbps and 161.1 Mbps in Europe rapidly increasing when the number of satellites increases. ²

This development comes with huge externalities: the emergence of a near-monopoly in space technology, light pollution due to the satellites, the risk of space collisions, and the ensuing multiplication of debris. The network of satellites requires an enormous amount of raw materials and energy, estimated to have equaled a world digital energy consumption of 3834 TWh in 2020, which is comparable to the 4077 TWh used in Europe as a whole in 2018. Furthermore, its share of world greenhouse gas emissions is estimated to be already 4% and is likely to double to 8% by 2025. ³

The third major development in the preparation for the 4IR is the construction of HDCs. The main management functions of data centers are to provide space for backups on a scheduled basis, store what is known as ‘the cloud’, regulate compliance to ensure access to those authorized and prohibit access to those not authorized; provide provisions to use applications, documentation, and printer facilities; the setup of a helpdesk, as well as the installation of monitoring, patching, remediation, and threat management systems. ⁴ One or more enterprises can store data, make use of the applications, and use the computing power in one HDC. If multiple companies/organizations make use of the facility, one talks of colocation data centers where the space and resources of the data center are rented out.

Until a couple of years ago, large organizations had their own data centers. Momentarily, developments go fast in building huge (hyperscale) data centers used by multiple companies. By the end of 2017, approximately just over eight million data centers were in operation worldwide, of which nearly 400 of HDC size. At the end of 2021, the number of data centers reduced just below eight million, but the number of HDCs increased to 700, thus only partly superseding the smaller data centers and rather being an extension.

HDCs contain cabling, switches, routers, and firewalls that mutually connect servers and connect them to the outside world. The HDC of Microsoft in Quincy, Washington State, for instance, comprises 24,000 km of fiber cable with millions of servers.

HDCs have externalities in the usage of high amounts of energy to cool the servers to a temperature below 25° Celsius, the use of fresh water in case the outside temperature rises above 25° Celsius, the storage of diesel fuel for the emergency generations used in case of power cuts, the noise produced when cooling the servers, and landscape effects due to their enormous size.

These externalities are not trifles. As of the end of 2021, 2.5% of all electricity used in the world is consumed by such hyperscale data centers. In Ireland, with 20 HDCs, these consume 30% of all national electricity. The externalities involved in an operational HDC of the size as was planned by Meta near Amsterdam are:

• Estimations from the company itself were that once the HDC would be in full operation, it uses over 2.3 million cubic meters of drinking water a year to cool its servers. 20% of this water evaporates during the cooling, and the remainder is filled with chemicals. The gross use of water in this data center is estimated to be equal to the average usage of over 13,000 households.

Considering these effects of the building blocks brings one back to basic principles like those established in the European Union (EU). In this case, the care for sustainability is crucial. Just to sketch the three basic options for governments, they can eagerly facilitate these facilities, prohibit them, or regulate their arrival.

South Korea and the Netherlands are seen as champions in their eagerness to welcome the facilities needed to establish the 4IR. Other governments seem to join them in welcoming the high-tech firms. Norway promised to lower taxes and slash red tape to attract HDCs. Saudi Arabia intends to invest 18 billion US $ to become the data center hub for the Middle East. South Africa is extremely proud to host Africa’s largest HDC (JOH1), expected to become fully operational in 2022.

As a recent book on ICT developments in South Korea shows (Chung, 2020), such eagerness is indicated by a national intervention strategy taking place at three levels:

1) Policies favorable for high tech companies;

Many rankings of countries on these three indicators are available. The European Digital City Index compares municipalities regarding their position as a start-up city and scale-up city. ⁵ The Tax Justice Network’s biennial ranking of tax havens, ranks countries according to their tax lenience. ⁶ Furthermore, some governments offer companies with stable and high consumption of energy, a volume correction on net rates going up to 90%.

The UN reports every year on the E-Government Participation Index, the e-Governance Development Index, the Online Service Index, and the Telecommunications Infrastructure Index (UNDESA, 2021: 119).

Third, concerning the promotion of an intelligent information society, rankings are available concerning ICT development ranking (ITU, 2022), Digital Competitive Index (IMD, 2021), and the network readiness index (NRI) in which indices are combined for Technology, People, Governance and Impact (Dutta and Lanvin, 2019: 13). ⁷ Although we as scholars often doubt the validity of such rankings, they are important location factors from the perspective of international companies.

Such offensive intervention has drawbacks as it disregards the externalities. It takes the tech firms at their word when they promise to be as efficient as possible and to be energy neutral, for instance, in 2030. Governments opting for this strategy also neglect that the tech firms depend on them perhaps even more than governments think their country depends on the presence of the tech firms.

Prohibiting such developments is the opposite of the above and implies installing protective barriers by imposing economic costs through, for instance, higher taxes, and legislation, aimed at curbing the expansion of high-tech companies, preserving the national identity, avoiding the high environmental costs, and protecting the country against wasteful investments that are contrary to its economic, political and environmental policies.

An example is China. This country banned cryptocurrency and mining for it through Blockchain technology, partly because of the extreme energy costs involved. 42 other countries have also put restrictions on the possibilities to deal with digital currencies. The externalities might not be the only reason. Other reasons for these bans are that these currencies pose an easy way to funnel money from and to illegal sources and that the widespread use of cryptocurrency might destabilize the national financial systems.

The main drawback of such an intervention is its ineffectiveness. In a globalized world, increasingly dependent on connectivity through the cloud and the internet, it is hardly possible to back off. Opting for this strategy runs the risk of giving the appearance that a country is similar to North Korea where the internet is only available for high-ranked officials, or similar to authoritarian countries like Kazakhstan, Belarus, and Myanmar in 2021, or Egypt, Libya, and Syria during the Arab Spring in 2011. These countries disconnected from the internet as part of a governmental policy to counter societal upheaval.

The third type of governmental intervention is to put restrictions on the new ICT developments. The building blocks are welcome if they meet the preconditions set by the government in terms of, among others, sustainability, but also in terms of citizen rights, and democracy, and economic fair play.

As mentioned above, in January 2022 the EU formulated eight principles to which the digital transformation should conform. ⁸ It is not the first time the EU tries to regulate the (wrong)doings of high-tech firms. In 2021, Facebook’s WhatsApp was fined for breaking the EU Data Privacy Law. Google was hit with fines totaling 8.25 billion euros ($10 billion). Amazon was charged with damaging retail competition, Microsoft over its abuse of its dominance over operating systems, and Qualcomm for abusing its monopoly position in computer chips.

Reinicke (2000) pointed to preconditions for regulatory intervention, and the need to close two gaps: the operational gap consisting of a lack of knowledge and information, and the participatory gap where intransparency inhibits decision-makers to make thought-through decisions (Reinicke, 2000: vii). The author called for global public policy networks consisting of individual and institutional actors to address these gaps. These networks need to be able to gather and disseminate knowledge, place important issues on the policy agenda, assist in the negotiation and settling of standards and regulations, and in designing implementation mechanisms for getting standards accepted (Reinicke, 2000: viii–xv).

The above cannot but result in the conclusion that public officials need to be trained in asking the right questions when confronted with high-tech companies. They need to be able to gather information, demand transparency, and conduct thorough impact assessments, be it in the form of cost-benefit analyses or evaluations. In the literature, these skills are known as skills in management information, in critical thinking, and problem-solving (Nene, 2018: 208; Van Laar et al., 2017), i.e. skills to be able to prepare for a thought-through decision-making process.

Regulating the changing decision-making by individual and corporate actors

From the literature, the idea might emerge that different elements of the 4IR are only relevant in specific areas of civil servants’ activity. e.g. nanotechnologies are expected to have a greater impact on the activities of officials in the field of industry, cryptocurrencies are expected to be more important for financial institutions, and fake news would only be relevant in the sector of culture and information. In practice, it is seen, however, that the impact of all these developments goes beyond such specific borders. All the elements of the 4IR are likely to have consequences for the public realm in general given the political, economic, socio-cultural, and sustainability issues involved.

Expectations regarding nanotechnology are that this will impact on scientific breakthroughs, increased industrial productivity through nanomanufacturing, and better healthcare, just to name a few shared expectations (Roco and Bainbridge, 2005: 1). We only have to use our fantasy and imagination to foresee the opportunities and dangers of the use of sub-molecular advancements in the army and in health care. Biomedical scholars, for instance, tell that personalized medication based on the unique DNA-chromosomes of individuals will become an efficient way to cure what are known at present as terminal diseases. Until now, millions of people take the same medicine because tests have shown these medicines are helpful for some of them. Nanotechnology promises the provision of personalized medicine, with a much greater chance of being effective and a much lesser chance of being harmful by reducing toxicity and off-target effects. For such reasons, the worldwide investments therein have exploded from 430 million US $ in 1997 to three billion US $ in 2003 to approximately 250 billion US $ in 2021 (Adiguzel, 2020, p. 1).

There are opportunities as well as risks involved. Because of the expectations on the one hand, and the potential risks on the other hand, the framing of nanotechnology varies from it being a key enabling technology to a highly contested technology as unemployment is lurking around, the position of countries and companies in markets changes, the inclusiveness in the access to the new opportunities offered is at stake, and political issues in terms of relative power and interests will emerge. For instance, for socioeconomic reasoning, the pharmaceutical industry is not that pleased with developments towards personalized medicine. It even is said that the interests in this industry pose one of the most important inhibitors for further developments in the medical use of nanotechnology (cf Hermann and Rosslein, 2015).

The emergence of Blockchain technology for decentralized but save transactions such as bitcoins and transactions between firms is a third development. It is not only of interest for civil servants working in the financial area, but has its own issues in terms of the large amounts of energy needed to secure safe transactions, mine for bitcoins, the volatility of the value of digital money, and the consequences for tackling crime as the technology is used by criminals and tax evaders.

The effects of the 4IR on individual and corporate decision-making are already partly visible and will become more apparent in the near future. One only has to reflect on the use of smartphones, to understand the major impact new technologies have on culture and lifestyle, consumer behavior, and information-gathering by individuals. Many have pointed out the possibilities, but also risks involved in the applications made possible by the confluence of the new technologies, in this case, the emergence of the IoT (Schwab, 2017, 2018). Among the risks are widespread misinformation, the use of bots, and hacking, made possible by the near-impossibility to distinguish misinformation from actual information. This guides individual choices in undesirable ways. The most recent report of the World Economic Forum on Global Risks mentions such cyber vulnerabilities among the 10 most urgent risks the world is facing in 2022 (WEF, 2022). This includes digital inequalities, cybersecurity, digital power concentration, and the risk of infrastructure breakdown. To acknowledge the new challenges emerging, one only has to consider the growth of deep fakes and “disinformation-for-hire” that are deepening mistrust (WEF, 2022: 49).

The IoT also has a serious economic impact on decisions on what goods and services to buy and where, how to organize one’s life and organization, and what goods and services to produce and where. The increasing number of options has consequences for the culture and lifestyles, economy, politics, and even democracy. The IoT allows ordinary citizens to buy something in China, India, and Europe irrespective of their home country. Corporations increasingly have the option to make use of robotics instead of manual labor with severe consequences for routine labor. Due to the possibilities of colocation of their management information, businesses are enabled to globalize further by adjusting their product cycle. Their headquarters could be in a country with tax havens, investments come from countries with much capital to invest, parts of the product are made in multiple other countries depending on labor costs, after which the final product is assembled in still another country having the technology for using micro-robots, and the marketing through social media is arranged in still another country (cf. De Vries, 2001: 394).

On the one hand, the increase of options to choose from seems desirable as it is indicative of the freedom of actors and results, at first sight, in an increased chance to have decisions resulting in optimal outcomes. Psychological research, however, shows that in reality, the opposite is likely to occur. The more options available, the less likely it becomes that actors choose the option with the most optimal outcomes and the less satisfied they often are with the choices made. Having too many options makes it difficult to obtain sufficient relevant information to judge each option, and to judge what is acceptable. It is therefore likely to result in dissatisfaction as actors continuously doubt whether other options would have been more attractive than the one they chose (Schwartz et al., 2002). Another drawback is that the number of options might result in actors becoming satisficers instead of optimizers (cf. Simon, 1955, 1959), asking not what would be optimal, but what seems good enough. Increasing the options might result in the substitution of a complex question into a simplified one. It might also result in more bias and noise in decision-making. ‘Bias’ implying that the distance between the actual decision made and the optimal decision increases, and ‘noise’ implying that the predictability of decisions decreases as the chance variability therein becomes larger dependent on personality, context, prejudice, low-quality information, the framing of problems to decide about, and what is called low decision-hygiene (Kahneman, 2017; Kahneman et al., 2021). In the 4IR skills to distinguish high-quality information from low-quality information (fake news) becomes crucial in preventing bias and chance variability in decisions.

Such bias and noise also occur in public decision-making. Due to artificial intelligence combined with big data, governments also witness an increase in the ways of possible steering in terms of dynamic steering, focused steering, personalized steering, and monitoring and steering based on the use of algorithms. One of the major developments is the use of algorithms, machine learning, and human-machine interaction when making decisions. Proponents tell us that algorithmic decision-making could result in more effective, efficient, personalized, and fairer decisions (Van Noordt and Misuraca, 2020). Opponents point to the potential injustice induced by these tools, the traps involved, and the potential hazards in terms of discrimination on the basis of race, ethnicity, gender, and socio-economic status (Ulbricht and Young, 2022: 5) Acknowledging the dangers, the EU has already made a General Data Protection Regulation ensuring that individuals will not be subjected to solely automated decision-making that has legal consequences or similar significant effects (Art. 22).

Nonetheless, many scholars see the benefits of government by algorithm, to establish a government that uses ICT systems to mine and analyze data, uses artificial intelligence to enable automated decisions about allocating resources, and optimally utilizing statistical models (Leadbeater, 2011: 18). Williamson similarly sees the desirability of algorithmic governance functioning through collecting, collating, and calculating data of citizens in order to predict their probable future needs (Williamson, 2014: 310). Kahneman (2017) even sees the use of algorithms as the ultimate remedy against bias and noise in decision-making, telling that “predictions and decisions generated by simple statistical algorithms are often more accurate than those made by experts, even when the experts have access to more information than the formulas use.” (Kahneman, 2017). ⁹ In practice, important decisions about people are increasingly made by algorithms (Kroll, 2015).

This observation in itself is already ample reason for discussing such changes in decision-making in PA programs and making students able to apply such techniques. Building algorithms is normally done on the basis of classification, regression techniques, and/or combinations of reasoned rules (Kahneman, 2017). Classification, like in cross-tabulation or statistical discriminant analysis, judges the likelihood of a case falling in one category or another. Dependent on a limited number of characteristics, i.e. reasoned rules, the algorithm can be used to advise the tax office to have certain tax returns checked for fraud or not; the police to increase or decrease surveillance in certain neighborhoods; traffic management in ways to regulate traffic; and a municipality to grant, amend or refuse permits. The range of possibilities for algorithmic decision-making is infinite. New Zealand is one example where automated systems are used for the purpose of ‘informing decision-makers to help solve complex issues that affect us all, such as crime and vulnerable children’ (Veale and Brass, 2019: 6).

Recent developments in the emerging 4IR show that in some countries the algorithms do more than just advice and pose a substitute for human decision-making. The decision-making itself transforms in automated decision-making. The idea is that socio-algorithms do better than humans, are more effective and efficient, enable mining of data for insights public professionals miss, remove organizational silos, and in general augment the analytical capabilities needed for decision-making (Ibid).

All this is not without risks. Classifications and reasoned rules resulting in automated algorithms may reinforce outdated practices because of changed societal values (Barocas and Selbst, 2016), reinforce discrimination (Kroll, 2015), and introduce their own bias (Kashin et al., 2015).

An example is an algorithm used until recently by the Dutch tax office. The tax office possesses big data - over the whole population - but with limited content (Name, address, age, (double) nationality, occupation, income, and whatever is filled in on the tax returns as deductible). The result was an algorithm based on parameters such as the family name (foreign origin or not), gifts to mosques mentioned in the tax return as legitimate tax deductibles, and income. If people had a double nationality, a low income, a family name pointing at foreign origin, and/or mentioned gifts to mosques as a deductible, the algorithm decided that they belong to the risk group, are to be blacklisted, continuously and extensively controlled, and not automatically entitled to social benefits, loans, grants, and permits.

When this algorithm became public, it caused justified public outcry over the racist and discriminatory character thereof. But the Dutch are not alone. An investigation of Big Data issued by President Obama in the USA in 2014, showed that big data technologies in general can cause societal harm beyond the damage to privacy, are plagued with an impenetrable set of algorithms with the danger of encoding discrimination in automated decisions (Jansen and Kuk, 2016: 373).

The use of algorithmic decision-making is full of many potential pitfalls (cf. Kuziemski and Misuraca, 2020: 4). Kuziemski and Misucara mention: the framing trap in which elements are included in the algorithm based on a one-sided and biased structuring of the problem; the portability trap in which it is assumed that if an algorithm is effective in one context it will also be effective in another context, while in fact, it might be harmful, or misleading when transferred over contexts; the formalism trap, in which the algorithm fails to account for all dimensions of a basic principle and, for instance, discards the difference between procedural and distributional fairness, or fails to appreciate different kinds of rationality; the ripple effect trap, implying that the use of an algorithm itself might impact on the values underlying behavior; and the solutionism trap in which the possibility of constructing and using an algorithm as such induces its use, according to the saying that ‘he that is good with a hammer tends to think everything is a nail’. This might be done even though it would be more effective and efficient to have decisions made without the technology.

Being skilled in building an algorithm that increases the effectiveness and efficiency of operations in government is certainly useful. This needs skills in classification and regression techniques as well as in constructing reasoned rules. To be skilled in evaluating such algorithms against national law, against the universal declaration of human rights, or the earlier mentioned principles as set out by the EU and to judge whether algorithms deserve a standing of being impartial and unambiguous (Gillespie, 2014) is in our opinion as important for public officials. This goes beyond being skilled in information management, i.e. being able to define, access, evaluate, and manage ICT information. It also goes beyond the need for critical thinking and problem-solving capacities. Rather, the ability to combine the three with regard to algorithm-based decision-making is essential to make optimal use of the opportunities and challenges posed by algorithmic governance. It is essential to train these skills in the classroom before graduates are appointed in the public sector as the goal needs to be a balanced view independent of political, economic, and ideological interests.