Do robots dream of electrical sheep?

Right to consent in AI as an ability power

American political philosopher John A. Simmons defines consent as an act by which one freely changes the existing structure of rights and obligations, typically by undertaking new obligations and authorizing others to act in ways which would otherwise have been impermissible for them.

In the context of the AI ethics lecture, consent is a power in this sense: "I can, by consenting, change the existing structure of rights and obligations, typically by undertaking new obligations and authorizing others to act in ways which would otherwise have been impermissible for them."

More perspectives of consent

From a deontological/rights perspective, people should never be treated merely as a means but always as an end. Consent is the most obvious way to ensure that we are respecting the moral agency of others, treating them as ‘ends in themselves,’ rather than merely as means.

From a utilitarian/consequentialist perspective, the greatest good for the greatest number. Obtaining consent has good consequences: people are happier if they have consented to initiatives that require their involvement; it’s easier to get people to engage, and so on. Here, consent is valuable as a means to an end, namely as a way of maximizing the utility of happiness.

Importance of consent

Based on an ex-Google engineer's adage "If you're not paying for something, you're not the customer: you're the product being sold," the importance of consent is highlighted as a critical factor distinguishing a user as a valued customer from being a commodified 'product'.

The core adage immediately frames the concept of consent. When consent is not robust, clear, and informed, users are not truly engaging as customers. Instead, their data, attention, and behavior become the 'product'" that is being packaged and sold to advertisers or other third parties — often without their full understanding or genuine agreement. True consent would mean agreeing to be a customer of a service, with clear terms of engagement.

Companies monetizing attention, rather than direct payment from users, are "incentivized to manipulate and spy on you." In this context, genuine, informed consent would ideally act as a crucial ethical and possibly legal barrier.

If users truly understood and agreed to the specific ways their data is collected, used, and how their behavior might be influenced, it would provide a check against companies' inherent economic incentives to exploit user information without explicit permission. Without consent, these manipulative and surveillance-heavy practices become the default.

The "incentives matter" theory suggests that companies paid directly by users aim to make them happy. It argues that this theory falls short in the 'real world', which is "a lot uglier" when companies monetize attention.

This ugliness implies that consent, if truly in place, would force the 'real world' behavior of these companies to align more with ethical user treatment, rather than purely profit-driven exploitation of their attention and data. The presence or absence of meaningful consent determines whether the user's well-being is genuinely prioritized or merely instrumentalized.

In the context of the adage above, consent is paramount because its absence allows companies driven by attention-based monetization to treat users as commodities, incentivizing pervasive surveillance and manipulation, thereby creating a less transparent and potentially harmful relationship.

Criteria for consent

Challenges with consent

Several challenges with professional consenting include:

1. Data scarcity: Text

2. Current information practices: Text

3. Anonymity: Text

Relevant consent frameworks

In Europe's GDPR consent principles:

Binding consent

Consent risk mitigation

Privacy by design principles:

Universal Usability principles:

Bonus: Indigenous data rights

Te mana raraunga: Māori data sovereignty

Chosen community members (of whānau, hapū, iwi) were entrusted with different kinds of knowledge. This creates obligations for these kaitiaki (stewards/guardians). Knowledge valuable to a community may be encoded in, for example, waiata (songs), oriori (chants) and whakairo (carvings).

Knowledge may be more or less noa (common knowledge) or tapu (restricted). The knowledge and the vehicles for knowledge that are socially and culturally valuable are taonga (treasure). Tikanga (customary practices) exist that inform how such knowledge is employed.

Potaua Biasiny-Tule: a tikanga Māori perspective on data

In the Māori language, ‘mātauranga’ derives from ‘mātau,’ the verb ‘to know’. ‘Mātauranga’ can be literally translated as ‘knowing’ or ‘knowledge’, but the term encompasses not only what is known, but how it is known — the way of perceiving and understanding the world, as well as the values or systems of thought that underpin those perceptions. ‘Mātauranga Māori’ therefore refers not only to Māori knowledge, but also to the Māori way of knowing.

To take a quote from the WAI 2522 report: "I do not see a difference between our whenua and our data. Our data has the same sort of connections as whenua does, but it is just a different format. I think that we need to protect our data and recognize that it can be collectively owned by hapū and iwi and whānau. No one individual can own our data or should own our data. There are also cultural considerations of the whakapapa of the data, the mauri of the data, what rights does the person who gave that data have, what rights do they actually give the person who is collecting it or the organization?"

“Our Pou” (Whare Hauora)

"Whare Hauora" translates to "house of health" in English. It is a Māori term that encompasses a holistic view of health and well-being, going beyond just physical health. Hauora acknowledges the interconnectedness of physical, mental, social, and spiritual well-being. It covers the topics:

Māori data and data sovereignty

Description

Principles:

Case study: Te Tiriti o Waitangi / Treaty of Waitangi

Current data ecosystem's respect to indigenous rights & interests

For starters, the New Zealand Census of Population and Dwellings area units and mesh blocks do not align with iwi or hapū boundaries. As a result, Census 2018 had a very low response rate for Māori. Iwi-affiliation data was of such poor quality that it was not published. From the lecturer, it does not help that Māori have little power in shaping official data that works for Māori to this day.