The Ethics of Google’s Use of User Data During COVID-19

Overview of the Google Article

On first viewing of Google’s COVID-19 Community Mobility Report for Australia (Google, 2020a), there does not appear to be an ethical dilemma, in so far as Google is concerned, related to user privacy concerns beyond the basic question of whether or not we should be freely allowing the large collection and storage of user’s data by governments or private industry. This lack of ethical concern is due to Google’s stated data collection methods.

The report outlines that the datasets used for the report were obtained by the same method that Google uses to collect, aggregate, and anonymise Google maps data. The report also states that this mapping data is only collected where user’s privacy settings are set to share such data; moreover, that this is an “opt-in” setting and is set to not share such data by default.

Key facts outlined in the report on data collection methods:

• Data is only used where a statistical significance is met.
• Data is aggregated and anonymised as per Google Maps guidelines/methods (Google, 2020b).
• Data is harvested from users, only where privacy settings allow for data collection – opt-in, off by default.
• User data is only used where the privacy threshold is met – must be enough users at a location to ensure anonymity.
• Differential privacy is used to add artificial noise to data, further preventing identification of any individual person.

Furthermore, there does not appear to be a secondary medical, nor personal use ethical dilemma present either as the report clearly states the expected limitations of use for the data provided.

The report states:

“This report shouldn’t be used for medical diagnostics, prognostics, or treatment purpose” and “isn’t intended to be used for guidance on personal travel plans” (Google, 2020a).

Therefore, any such use goes against the explicit recommendation of the report and outside of the stated use boundary.

However, an issue does appear to arise when considering the alleged claim that Taiwan is using the same dataset source to create “electronic fences” around quarantined citizens to monitor their movements (as raised in the assignment brief, citing The Verge, 2020). This appears to go against Google’s recommendations to not use the data for medical prognostics. In this case, the data is being used to monitor and mitigate the pandemic spread of a viral contagion. Moreover, it also appears to contradict Google’s own statement that users cannot be identified individually from within the dataset, as outlined in the data-collection facts above. If Taiwan’s alleged use of Google’s user data is true, it raises further questions against Google and the Google report; specifically, the opt-in nature of the data collection.

Under normal circumstances, the above privacy concerns alone would be enough for further ethical and legal investigation. However, we are not currently operating under normal circumstances, but are actively experiencing a world-wide viral epidemic. Understandably, with this will come some latitude to how we approach individual privacy and autonomy in relation to the greater public health. There should always be a concern and a sense of reticence when considering giving up any freedoms, privacy, or rights; as once the cat is out of the bag, it is very hard to put it back in. There is always a risk of bad faith actors taking advantage of compromised people and circumstances to further their own wealth, power, and interests, whether that be individuals, private industry, or government.

Thus, taking all the above into account, the main ethical issue stands to be Google’s apparent conflicting position on user privacy and the possible use of non-anonymised user data without user consent. Which raises the generalised ethical question: Does the current pandemic situation allow for such ethical slipperiness in relation to the use of non-anonymised user data for the greater good?

An Ethical Analysis

In this section I will apply two ethical theories, Mill’s utilitarianism (Brink, 2018) and Kant’s deontological moral theory of The Categorical Imperative (Johnson & Cureton, 2019), to the problem: Does the current pandemic situation allow for such ethical slipperiness in relation to the use of non-anonymised user data for the greater good?

I have chosen to use the generalised form of the ethical dilemma as the Kantian analysis is better suited to individual’s and their actions, rather than organisation and governmental agencies and their respective actions. Moreover, as indicated by the conflicts within the Google report, Google itself may be acting against its own ethical guidelines in its joint actions with Taiwan. And regarding Taiwan, the analysis that follows will be considered on the individual user and their respective user-data, which will account for the Taiwanese populace.

The Kantian Analysis

According to Rachels & Rachels (1986), Kant’s ethics are built from the idea that human beings, being the only rational animal, hold a unique and special place within the world. And that because of this rationality are above all other species, above all ends, and are never to be used as means to an end. Moreover, as the focus is on the value of humans, and specifically their rational agency, his ethics are primarily interested in the intentions and motivations guiding one’s actions, rather than the consequences of one’s actions.

Central to Kant’s ethics is The Categorical Imperative, which he describes in three distinct formulations (Kant, 2008, p. 34):

1. “Act as though the maxim of your action were to become, through your will, a universal law of nature.”
2. “Act in such a way as to treat humanity, whether in your own person or in that of anyone else, always as an end and never merely as a means.”
3. “Act only so that your will could regard itself as giving universal law through its maxim.”

Exploring Kant’s ethics in completeness is far beyond the scope of this assignment; for conciseness, I am going to limit this analysis to Kant’s first formulation of his Categorical Imperative, also known as “The Formula of Universal Law” (California State University, 2020).

Applying the contradiction in conception test to the third maxim, we must ask ourselves: If in every public health crisis, every person always sacrifices their privacy in order to save the lives of other, would this always lead to the saving of people’s lives?

Unfortunately, this is a difficult question to answers as it relies on many factors that are hard, even impossible to know in detail or in advance of the question. However, we can assume, based on the Google report, that in this case private data is relevant and is at least suspect of being beneficial to public health. So, for the sake of this analysis, we will assume that it does indeed always lead to saving of people’s lives and thus passes the contradiction in conception test.

Applying the contradiction in the will test, we must ask ourselves: If in every public health crisis, every person always sacrifices their privacy in order to save the lives of other, would this conflict with a person’s individual will?

Thankfully, this is a much easier question to answers. If the third maxim were a natural law, it would most definitely conflict with a person’s individual will as a right to privacy as prescribed in the Universal Declaration of Human Rights; where it states: “No one shall be subjected to arbitrary interference with his privacy, family, home or correspondence, nor to attacks upon his honour and reputation. Everyone has the right to the protection of the law against such interference or attacks.” (United Nations, 2020, article 12). In a world where it was not a choice to refuse such privacy intrusions, such a right would be nullified, and thus fails the contradiction in the will test.

In following Kant’s Formula, that we should act only on that maxim through which you can at the same time will that it should become a universal law, in regards to the current ethical dilemma, we would then have the imperative to not act in accordance.

The Utilitarian Analysis

Utilitarianism can be summed up simply as the view that “the morally right action is the one that produces the most good” (Driver, 2014). It is a form of consequentialism, which is a class of normative ethic theories in which the moral value of one’s actions is predicated on the consequences of said actions. This contrasts with Kantian ethics, where the ultimate outcome is not relevant to the moral question of right or wrong, but rather the moral focus is directed towards the intention motivating the act. Moreover, there are various versions of utilitarianism, with differing influences ranging from the theological to the Epicurean, each having their own description of what is ‘good’ and its antithesis, the ‘bad’. For the sake of this discussion, I will generalise this antithetical to the form ‘betterment vs harm’.

To fully grasp the implications borne out from the simple outline of utilitarianism as described above, a few qualifiers are necessary.

The Utilitarian qualifiers (Hospers, 1972, pp. 4-8):

1. An act must be voluntary, in so much as there must be a choice to act otherwise.
2. Time has no value on betterment – it does not matter if the effect is now or later.
3. Betterment is not to be considered alone; must also consider harm.
4. Just because an act produces greater betterment than harm, does not itself make the choice right; there must be no other choice that could have been made that would lead to greater betterment and/or less harm.
5. Cannot assume a simple calculus will always be between betterment and harm; it is just as likely to be a choice between two choices of relative harm.
6. The Utilitarian calculus must be made impartially – all actors, including oneself, are weighted equally.
7. To the utilitarian, no law, in and of itself, is sacrosanct – the utilitarian follows their calculus wholeheartedly and without weight towards existing laws or commonly held beliefs.

With these guiding principles, the utilitarian is both motivated and able to evaluate ethical and moral choices in terms of betterment vs harm, for the many vs the individual, reasonably and impartially.

Regarding steps 3 and 4: Evaluating a given consequences respective probability and utility to within any degree of accuracy is difficult. And while there may be relevant statistical and utility projection methodologies that are beneficial to this end, they are beyond the scope of this discussion. Thus, for this analysis I will simply apply an estimated general best-fit value for each respective option, such that:

• Allowing the use of personal user data helps many in some cases and hurts individuals in most cases.
• Preventing the use of personal user data hurts many in few cases and benefits individuals in most cases.

Using the table above, we can identify the best utilitarian prospect by applying a metric to the probability and utility columns, such that: Very Low = 1, Low = 2, Med = 3, High = 4, Very High = 5.

Using this metric gives us:

A

(5 + 5) + (4 + 1) = 15

B

(3 + 2) + (2 + 5) = 12

While it is relatively close between the two options, there is a slightly better outcome for overall betterment for A over B. This is mostly due to both the benefits and high probability for all to benefit from using user data to slow the spread of contagion during a viral pandemic. Thus, consequently, the actions of A is the utilitarian’s recommended choice.

However, there is a caveat: one of the main problems for any utilitarian calculation, and one that is apparent here, is the difficulty to ascribe probability and utility values for a given act. Hence why this cannot be considered necessarily the morally right choice; given more information and a better metric, a different and possibly better outcome is very likely.

Summary

Google reports that its data collection methods are both anonymised and optional for the user. They also state the limited use range of the data specific to the generated report of user activity during the Covid-19 isolation period; specifying that it should not be used for medical diagnostics, prognostics, or treatment purpose, and further, that is not intended to be used for guidance on personal travel plans.

However, Taiwan’s apparent ability to use Google’s data to both monitor and “electronically fence” quarantined citizens raises questions of Google’s credibility and introduces doubt regarding their officially stated position of user anonymity, along with an apparent conflict with their own data use policies. This raises the ethical question of whether the current Covid-19 pandemic situation allows for such ethical slipperiness in relation to the use of non-anonymised user data for the greater good? I then turned to two distinct forms of philosophical ethics attempting to analyse and answer this ethical dilemma, Kantian Ethics and Utilitarianism

Given the distinct and separate approaches to ethical analysis as presented by Kantian and Utilitarian ethics, it’s not surprising to find that their respective calculus led to opposing choices: Kant’s approach leading to choose not to share data; whereas the Utilitarian approach leads to choosing to share data. I think a major reason for this difference is due to their differing primary moral focus: Kant’s valuation of motivations and intentions over outcomes; conversely, utilitarianism valuing outcomes most of all. It is these fundamental differences that appear to push the ethical frameworks into opposition as to what is the right choice, at least in this situation.

One further thought, this ethical question was analysed in the generalised form and directed towards the individual and their respective choices. However, this may be moot when considering Google’s possible breaking of its own privacy guidelines, which raises doubts against their privacy policies, specifically their ‘opt-in’ data-collection policy. Regardless of choice, users may, in fact, be powerless to act on who controls and has access to their private user-data, moving the analysis from the personal ethical dilemma of users to the unethical acts by organisations.

Lastly, on self-reflection, if I were to apply my own ethical intuition to the case in hand, I find myself leaning towards the Kantian choice as the risk to personal liberty by loss of privacy seems far greater in the long term then the short term benefits of giving them up. However, like the utilitarian, I too would feel the need to apply the utilitarian qualifier that there may be a better, more morally right choice to be made beyond the choices presented. That with further knowledge of the moral landscape and the choices therein my response to this dilemma may change. Consequently, the morally right choice appears a moving target; one that is aimed for but not necessarily ever caught.

References