Data Driven versus Value Based

Hannah Arendt

It has taken me some time to read Hannah Arendt's work. But in my talks with Bram Abels, he regularly cited from her books to help clarify the issues we were talking about. Her ideas witnessed a unique and clear cut world view, as well as a strong logical structure of her analysis and views. So I read her work, and listened to her recorded speeches, in the beginning mainly to understand the mind and actions of a 'totalitarian', urged by the horrific invasion and genocide in Ukraine.

I also followed the recordings of the Virtual Reading Group led by Director Roger Berkowitz for the Hannah Arendt Center for Politics and Humanities at Bard College, which I recommend, because of the in depth analysis of Berkowitz, and his systematic and analytical study of Arendt's book The Origins of Totalitarianism:

Dutch Childcare Benefits Scandal

Suddenly it struck me that I had recently witnessed the patterns Arendt describes in The Origins of Totalitarianism somewhere else; the so called: Dutch childcare benefits scandal. The scandal resulted in the resignation of the PM and his entire Cabinet. The scandal has a dedicated Wikipedia page that provides a practical historic overview.

See also the Politico article: Dutch scandal serves as a warning for Europe over risks of using algorithms . In essence: since 2013, in the Netherlands, 26,000 innocent families were wrongly accused of social benefits fraud partially due to a discriminatory algorithm, used by the Dutch tax authorities. Citizens were forced to pay back money they didn’t owe, many families were driven to financial ruin, and some were torn apart. Others were left with lasting mental health issues; people of color were disproportionately the victims. A remarkably large number of children (1115 [in the period: 2015 to 2021]) from families affected by the benefits scandal, were removed from their homes.?

Update: on May 11 2022, Statistics Netherlands (CBS) published updated data; the total of out-of-home placements of children (2015 to 2021) was: 1,675.

A parliamentary?report ?into the child care benefits scandal found several grave shortcomings, including institutional biases and authorities hiding information or misleading the parliament about the facts.?

Aim of this post

In this contribution, I shall follow Arendt's analysis of The Origins of Totalitarianism and use as a case the Dutch Childcare Benefits Scandal, in order to make some observations, with regards to certain assumptions and implicit logic underlying our strategies to make decision, and more specifically, when we use algorithms to support decision making.

My main conclusion will be that we can and should learn from Arendt's The Origins of Totalitarianism, to prevent unwanted and unforeseen outcomes of questionable decision making strategies that have unprecedented negative effects on the citizens affected by the 'system of decision making', not only in terms of financial injustice and health problems, but more fundamentally, the loss of their 'legal and moral person'. All this as a result of a rather arbitrary process of de facto pointing out random victims by a system operated by bureaucrats on behalf of a democratic government, that causes the targeted citizens to loose their fundamental freedoms and rights; they become victims of state propaganda: these 'fraud perpetrators' were forced to pay back money they didn’t owe, often resulting in financial ruin, and in many cases have their children taken away from them.

Line of reasoning in 'The Origins of Totalitarianism'

1. Totalitarianism differs essentially from other forms of political oppression known to us such as despotism, tyranny and dictatorship. [page 939, Chapter 13]
2. It is in the very nature of totalitarian regimes to demand unlimited power. Such power can only be secured if literally all men, without a single exception, are reliably dominated in every aspect of their life. [page 930, Chapter 12]
3. Totalitarianism strives not toward despotic rule over men, but toward a system in which men are superfluous. [page 931, Chapter 12] Even leaders are replaced with startling ease and forgotten with startling swiftness. [page 659, Chapter 10]
4. Totalitarian movements are mass organizations of atomized, isolated individuals. Compared with all other parties and movements, their most conspicuous external characteristic is their demand for total, unrestricted, unconditional, and unalterable loyalty of the individual member. [page 691, Chapter 10]
5. Such loyalty can be expected only from the completely isolated human being who, without any other social ties to family, friends, comrades, or even mere acquaintances, derives his sense of having a place in the world only from his belonging to a movement, his membership in the party. [page 550, Chapter 10]

So, the totalitarian ruler, deconstructs the nation-state, with its classes, into a movement of the mass of atomized, isolated, superfluous individuals, with the use of propaganda, 'outrageously insulting common sense' [page 744, Chapter 11] and terror, for instance via the instrument of a secret police, 'so eager to convince its victims of their guilt for crimes they never committed, and in many instances were in no position to commit' [page 775, Chapter 11].

Arendt continues to describe the 'radical evil' [page 743, Chapter 12] tactics by the totalitarian murderers. After the individual is lost, the social status and the protection of the social class, the atomized 'mass individual' is stripped of rights and citizenship; 'the destruction of a man's rights, the killing of the juridical person in him, is a prerequisite for dominating him entirely' [page 920, Chapter 12]. 'The next decisive step is the murder of the moral person in man' [page 921, Chapter 12]. This is achieved by the deportation of the Jews, the homosexuals, the criminals, political opponents, and potentially every inhabitant of the totalitarian state [page 730, Chapter 12] to the concentration camps, the 'death factories' [page 743, Chapter 12], or 'veritable holes of oblivion' [ibidem] where people just disappear. The victims are not 'human' in the eyes of their executioners [page 743, Chapter 12]. 'After murder of the moral person and annihilation of the juridical person, the destruction of the individuality is almost always successful' [page 921, Chapter 12]].

Arendt concludes: 'This is the real triumph of the system: "The triumph of the SS demands that the tortured victim allow himself to be led to the noose without protesting, that he renounce and abandon himself to the point of ceasing to affirm his identity." (...) For to destroy individuality is to destroy spontaneity, man's power to begin something new out of his own resources, something that cannot be explained on the basis of reactions to environment and events. Nothing then remains but ghastly marionettes with human faces.' [page 736, Chapter 12]

The Loss of the Legal Person

In light of the horrific observations by Hannah Arendt, and many consequences thereof, the case of the Dutch childcare benefits scandal, can be seen in a different, inconvenient, light.

Because here too, citizens were rather randomly victimized, singled out, and were made to believe, by the Dutch tax authorities, that they had committed fraud, when in reality they had not. Consequently rights were taken away, in many cases the kids were removed from their homes. Parents broke up. Went into hiding - a different hole of oblivion - and in many cases disappeared from society and more importantly: hid for their government and the bureaucrats. The irony, is that this was not part of a deliberate or secret plan to do harm to the tens of thousands of Dutch parents, and limit their rights and freedoms.

There are many factors that led to the painful outcome for the victimized parents.

From a Privacy Regulator Perspective

The Dutch Data Protection Authority’s investigation (DPA) decided that the Benefits Office of the Dutch Tax Administration should not have processed the (dual) nationality of childcare benefit applicants in the way it did for many years. According to the results of the DPA's investigation, this practice was unlawful and discriminatory. The Benefits Office also processed the nationality data of childcare benefit applicants for the purpose of combating organised fraud, even though this data was not necessary for this purpose.

Lastly, the Tax Administration used applicants’ nationality (Dutch/not Dutch) as an indicator in a system that automatically designated certain applications as risky. The data was not necessary for this purpose either.

It is unlawful to use nationality data to assess applications, combat fraud and determine risk. In other words, the Tax Administration was not allowed to do what it did.

On December 8 2021 the DPA has imposed a €2.75 million fine on the Minister of Finance because he is responsible for the processing of personal data within the Tax Administration.

From the Council of State Perspective

See also the Raad van Staate's Reflection Report [Dutch].

The Council of State, the highest Dutch administrative court, generally ruled in favour of the Tax Authorities, and not the parents, with regards to the strict application of the 'all-or-nothing' settlement with regard to childcare allowance: if an applicant could not prove that he was entitled to the full amount of allowances, the parent had to repay the entire advance. Until the autumn of 2019, the Council of State maintained this working method, despite signals that the law could be applied less strictly.

The Council of State spokesperson, chair of the relevant department offered apologies on behalf of the Council of State, stating: 'We should have provided better legal protection.'.

From the Parliamentary Interrogation Committee on Childcare Benefits Perspective

See the report [Dutch].

A parent's mistake in a Childcare Benefits application was seen as fraud, causing parents to be unjustified stigmatized as fraudsters. The Parliamentary Interrogation Committee does not only accuse the Benefits Office / Tax Authorities - but also the legislator and the judiciary.

It also becomes clear in the Report [page 40] that in 2011, the anti-fraud policy was further tightened on the basis of the coalition agreement. One could state, based on this, that the government was treating citizens from a perspective of distrust.

From a 'data' perspective

Many people will say things like: "I work data driven". Or: "the data shows that ...". I think this attitude is the cause for many bad decisions. Where not the relevant question is leading, but whatever data may be considered to be conveniently available, and somehow related to the question at hand. This 'data dominant' perspective will encourage people to combine datasets that are not relevant for the purpose, and thus illegal to process for that purpose. One of the GDPR principles for data processing is data quality, and this was not ensured, based on the DPA's findings, is my interpretation. People will more quickly identify the GDPR principle of purpose limitation, but if data quality is not ensured, then the outcomes are unreliable, and have, as we have seen, a negative impact on the rights and freedoms of the victimized parents.

The alternative is to work 'value driven'. What is the responsibility the government has towards its citizens? How can citizens trust their government to protect them, their data, their rights and freedoms? What values are leading for the processes of the Benefits Office, of the consistency of the legislation to be executed by the Tax Authorities? What balances need to be struck and how, and by whom? Is an informed human decision built in the process, when the balancing of interests is to be done? Can the government explain how and why decisions are made, based on which evidence based indicators, relevant for the context?

In a previous role, the hype at the time was called 'outsourcing'. I have had my fair share of experience with outsourcing with partners in India and the Philippines. Many things went really wrong for a number of reasons, cultural aspects, effecting communication and expectations. But for me the best lesson learned was after the fact, when I saw a maturity model for outsourcing. It basically stated that, given your organization's maturity and knowledge of the processes to be outsourced, you would do well to start outsourcing the things you already know fully and really well and you have documented scripts for. Say a service desk for common incidents. When the organization has reached the next level of maturity, one could think about outsourcing well-documented tasks in 'not so complex projects'. The next level of maturity, would justify thinking about doing innovation projects with your outsource partners.

The same goes for applying Machine Learning and self learning algorithms, I propose. This is in fact also a way of outsourcing. Yet, how can one outsource the responsibility of decision making, if there is not enough organizational maturity on relevant, evidence based, indicators, to a self learning algorithm? Constructing crappy profiles, which are copied and used in other contexts also, will lead to unreliable outcomes and at the same time can have real life negative effects for the individuals concerned. How can the Benefits Office errors be made undone? Especially since the 'fraudster' data and their details were shared across government departments? Data processing is evermore complex, with linked datasets, in chains - and who is responsible for the whole chain of processing?

In cases of identity theft we see that victims keep being harassed by governments, for long periods of time, after it is clear that someone was victim of identity theft - also because the 'criminal records' often are shared with other countries too, who have not deleted these records yet.

Epilogue 1

Reflecting on the data driven and the value based approach, I can not help but hear Immanuel Kant, with his: "Gedanken ohne Inhalt sind leer, Anschauungen ohne Begriffe sind blind" / “Thoughts without content are empty, intuitions without concepts are blind”, in his Kritik der reinen Vernunft.

It was Kant who studied and examined the way we can make decisions at all (Kritik der reinen Urteilskraft) as well as the one who coined the term 'radical evil' that Hannah Arendt picked up for her Totalitarianism work. I plan to dive deeper in the next weeks into lessons to be learned from Kant, who asked these four questions (Lectures on Logic):

(1) What can I know? (2) What ought I to do? (3) What may I hope for? (4) What is man?

As Paul Nemitz revealed some time ago , the GDPR has an underlying Kantian Ethics, so there are many relevant reasons to, again, read Kant. So, not only for his categorical imperative (German: kategorischer Imperativ), but also for his epistemology, logic and judgement, and their inherent interplay.

I thought it very revealing that just recently the book: Kant and Artificial Intelligence was published (open access) by De Gruyter. Looking into this as well.

Epilogue 2

In the aftermath of this long read, it is also helpful to bring in Nick Bostrom's parable: 'The Unfinished Fable of the Sparrows' in his book: Superintelligence. Paths, Dangers, Strategies (2014).

Perhaps you have read the book, perhaps not. Parker Harper has made a visualisation of the parable, authorised by Nick Bostrom:

See also

For purposes of balance, I would like to invite you to check this edition of the series: ‘People also Ask...and we answer’ by Clearbox AI , with the marvelous Shalini Kurapati (CEO) and Alessandra Nicolosi, where we discuss also the good of AI, and how privacy and AI are related. I applaud the transparent way of working of Clearbox AI, their focus on data quality, and their use of synthetic data.