Moral licensing, AI teams, and you - a problem yet to be reckoned with
- Summary:
- Most of us have heard of cognitive bias - and the problem it can pose. Less well known is the practice of moral licensing. But it's an issue AI teams need to consider.
Suppose you had the opportunity to purchase some product at the grocery store, and there were two competing products to choose from. One was the store's brand, and the other was listed as no antibiotics, no GMO, etc. The second one seemed to be the same in every respect except the price, which was 35% higher.
You chose the second because you felt a responsibility to "save the planet." When you checked out, you put a Lindt chocolate bar on the belt but you didn't feel guilty, even though you'd previously decided to give up sweet treats.
There is a name for this unconscious balancing - moral licensing.
Moral licensing is a cognitive bias that allows people to act "immorally" without affecting their moral self-image. These terms of moral and immoral are pretty broad, as I'll explain below. The effect of moral licensing is to dampen the impact on one's self-image of performing a relatively immoral action (perhaps just self-indulgent), having previously performed a moral act.
Moral licensing isn't always unconscious. If it is intentional, it can be an attempt to soothe one's conscience in a deceptive effort to avoid criticism. Moral licensing is not always a one-to-one effect either. One may feel they've banked enough moral actions to compensate for something on a grander scale.
In the above example, indulging in a chocolate bar will not change the world, but unfortunately, moral licensing has pronounced and dangerous effects at other levels.
In the AI Ethics field, we seem to focus on the data and training of models, but not how the various players make their decisions. The graphic below depicts that many people are involved in producing an AI application, and how each one makes decisions is critical. Each one is likely to engage in moral licensing in their decisions, and it's not often evaluated. Moral licensing is a vital component that we as an industry have overlooked.
AI team structure graphic via TalentSeer's 2020 AI Talent Report
Moral licensing in the customer journey
In a mostly benign way, moral licensing can be a useful technique in, for example, up-sell/cross-sell situations. If a consumer is presented with a decision that has a positive moral aspect, the consumer may be more likely to agree to a relatively immoral decision subsequently. It's the sequence that matters because the first decision influences the second. Online retailers can stage the customer journey options, with different product categories presented to consumers.
Moral licensing in organizations
I've written several times about the "for good" movement, how companies gain publicity engaging in acts of a philanthropic gesture, or, more frequently, participatory donations with customers for some "good cause." Moral licensing is at play in most of these programs, particularly corporate social responsibility (CSR).
Amazon: Shop at the site smile.amazon.com and select a charity. You are notified that a portion of your sale or purchase will be sent to this charity in subsequent visits. If moral licensing is useful, I might make an impulsive purchase (I tried this, and I purchased a few books that I might not have).
Another example is how CSR activities partly shape consumers' opinion of companies. A common (and inexpensive way) is to solicit donations for worthy causes. Airlines did this before the COVID-19 epidemic crushed them. It was common to see requests for donations (matched, of course) to a charity like St. Jude Children's Research Hospital, asking customers to donate at the beginning of an itinerary and fare search. Donation options are offered before the actual purchase and payment for a ticket. The moral licensing effect is evident by the contrast between the self-indulgent and the morally positive decision situation. The customer may choose a more upscale hotel or luxury rental car.
This is innocent enough as the "immoral" action is the indulgence of the customer. A darker version of this is when the company engages in CSR activities to mollify their collective self-image while their business is clearly immoral. One of the most egregious examples is Monsanto. The Monsanto Fund accepts grant proposals (for programs outside of the U.S. only) providing basic education support to improve education in farming communities, and meeting critical needs in communities by supporting food security, sanitation, access to clean water, public safety, and various other local requirements.
What is the immoral part? Monsanto is a perfect example. In one word - GMO. Not only haven't they lived up to their promise, but they've also actually raised the cost of farming, require more chemicals, and they lock farmers in. From the beginning of agriculture, farmers have saved seeds from their crops to plant the next year. When farmers purchase GMO seeds, they enter into contracts with seed companies and sign an agreement to purchase new seeds each year and not save seed from their crops to plant the following year.
This is also a litany of tobacco-company-like lies about their 'products' safety, particularly herbicides like Roundup, which used to be called Agent Orange.
Cultural difference in moral licensing
Marketing efforts that employ moral licensing need to be aware of cross-cultural differences. Marketing measures driven by moral licensing can expect to see varied results in different countries.
Research "shows that the effect of moral licensing is stronger in North America than Western Europe, and the moral licensing effect occurs in a reversed direction in Southeast Asia."
I'm not completely satisfied with that perception. There are approximately 580 million people in North America and about 750 million in Europe. There is a lot of diversity there. If the researchers meant North America as the U.S. and Canada and some small Caribbean countries, then at 365 million and Europe as Western Europe, at about 200 million, perhaps that makes more sense. There are strong shared historical roots and cultural, political, and economic bonds between Western Europeans and North Americans, and likely they share many beliefs and values.
There is substantial variance in how they view morality, which affects their conceptions about moral or immoral behavior. According to a study by Pew Research Center 2014, there is a wide disparity in religion's effect on their life, with North Americans giving much more value to religious beliefs than Western Europeans.
Incorporating unselfishness as a key personality trait is valued much higher in North America than in Western Europe, according to this (Pew) research. "Being a good person is more important to North Americans than Europeans, which may explain the moral licensing effect being more potent in North America." If morality issues are seen as far more pivotal, more North Americans will view certain moral flexibility in decisions as undesirable than Western Europeans. Vice versa, Western Europeans are more likely to allow some moral flexibility. That is, if you accept this finding, which I find not only doubtful but so general, it's hard to place any value on it. Their logic implies the value of a license that encourages an individual to behave immorally is less a factor in North American culture than in Western European culture.
My take
I'm not entirely convinced that every decision I make is a yin-yang between moral and immoral. It seems too contrived, too simplified. However, we as an industry, attempting to pursue trustworthy and ethical AI, have not examined the whole thought process of the decisions the numerous actors make in developing an AI application, and I do believe that moral licensing plays a role.