Product transparency, Corporate experiments

by guy keulemans on June 22, 2015

A few days ago Michelle N. Meyer and Christopher Chabris wrote an article in The New York Times titled “Please, Corporations, Experiment on Us.” The article raises the issue of corporations conducting experiments on their customers, users and employees – as seen last year when Facebook admitted to manipulating its data feeds to elicit emotional responses from a group of its users, in comparison to other Facebook users. The dating site Ok Cupid conducted a similar experiment on a small group of its users, manipulating its compatibility algorithm to test that algorithms usefulness. The article makes the case that while such corporate experiments were condemned, they can nonetheless be useful and should be encouraged. I find that argument problematic, perhaps even dangerous, though to be fair it not suggested that corporate experimentation should be done without ethical standards, and this is discussed more explicitly in the linked academic paper written by Meyer. However, what I think is most interesting about the article is that it has another implication, which is that ethical standards need to be asserted for a broader range of corporate activities, and specifically the production of consumer goods.

The authors set up the premise that corporate experiments like Facebook’s or Ok Cupid’s are ‘A/B’ type experiments (comparing one set of user/consumer effects ‘A’ with another set ‘B’, the control). But corporate activities, such as releasing a product or making system-wide changes to software algorithms also produce effects on user/consumers in the form of A “imposed by itself”, that is, not compared to anything properly known, quantified or assessed (i.e there is no control).

The authors don’t define that latter practice, but I’m going to call it ‘A/{ }’, where { } means an empty set, though the existing set of conditions to which it refers are not actually empty, just unknown.

The claim of the article is that because we readily accept products into the market in the form of A/{ } (and without the level of consent or ethical standard required in scientific studies), we should also accept the practice of producers manipulating a product for one group in comparison to another in the form of A/B, as a means to gain better experimental data.

I think the mistake in the article is that the significance of this is not that the first practice A/{ } legitimises the second practice A/B. Rather, A/B exposes a problem with A/{ }, and that is the problem of consent and disclosure.

In regard to this, the article notes that the ethical problem with the Ok Cupid and Facebook studies are that they did not disclose the experiments, obtain consent or provide a way for users to give consent. Users were not informed to the extent they could opt out and be in the ‘control’ group, or leave the platforms altogether.

This guides us to consider the issue of consent for A/{ } activities. Consumers and users should, in theory, have such consent because they can choose, or choose not, to buy or use a product. The problem with this consent, however, is that it is not truly informed consent, because producers do not disclose all the information about their products that consumers, arguably, need to know. There are a very many standards that already apply to consumer goods – standards of safety, for example, so that electronic products don’t shock or explode, or so chairs don’t collapse. But what is not disclosed? With increasing levels of blackboxing, consumers may not know how to access, modify or repair some products. Furthermore, there is much information about supply chains and their social or environmental relations that are not disclosed. For example, we don’t know very well how many blood metals from Congolese warzones are in our Apple phones, or how many Chinese factory workers killed themselves are a result of producing it, or how difficult and polluting the phone is to discard or recycle. Likewise, we don’t know the intricacies and effects of Facebook’s sharing algorithms on our mental health or consumer behaviour. This is because these corporations choose not to disclose it. It may be what they consider unpublished intellectual property or it may be competitive business information unappealing to customers.

Its possible that manufacturers don’t disclose some information because they don’t know it themselves, but historically there are famous cases in which very useful information was deliberately suppressed or hidden from consumers. The health risks of tobacco and high-sugar diet foods are two notable examples.

With better transparency, consumers should be better able to choose to stop smoking or reduce their sugar intake, or buy more ethically produced phones or use an alternative, less manipulative social media platform.

Transparency and consent concerns employment also. In the hypothetical example the authors give, a CEO implements an action to bolster employee savings. Its an interesting hypothetical because the action is itself the disclosure of that action (the release of information regarding the saving behaviour of their peers). This complicates consent and the capacity for an employee to opt out within the same corporation. However, if that action is fully disclosed to all employees and likewise, but different, actions are fully disclosed at another corporation, then the employee can use that information to consider the benefits of changing employers. In turn, market effects should provide the corporation with information needed to create beneficial employment practices and attract better employees. If we set the goal as not achieving just the corporation wants directly, but achieving that which helps individuals, societies and corporations together. (Further to that, it could be asked, who is making the claim that it is beneficial for employees to save more income anyway, and why? I would be suspicious of a CEO that claims a policy is for an employee’s benefit, but feels that the policy possibly shouldn’t be disclosed in order for them to benefit – even if that’s an unknown at the time of non-disclosure.)

The article argues that corporations should be encouraged to experiment more. This doesn’t necessarily imply they can do so without ethical standard, and the linked academic paper makes it clear that this is not the case, but the danger is that the article may be interpreted that way.

It makes sense to argue that it is ok for corporations or producers to conduct such A/B studies, but only if they are are subjected to well established ethical standards or for research, such as those used by universities. In the US and Australia, these are rigorous.

The subsequent and more compelling extension of that argument is, however, that ethical standards should apply to the manufacture or modification of all products – the A/{ } experiments. What I believe we should be considering is how ethical standards of transparency, disclosure and consent can apply to the design, manufacturing and marketing of products across the broader realms of production and consumption, so that consumers have a better quality of informed consent in general.

Its readily accepted that the specific health-related information and effects of tobacco and food should be disclosed to consumers at the point of sale and as packaging information. Why not the specific socio-environmental related information and effects of consumer goods?

Leave a Comment

Previous post:

Next post: