Discussing the ethical implications of AI and how to regulate them has been the central topic of our conference for the past 4 years, but we have never really defined what we mean by ethics! In our last online conference, we met Olivia Gambelin, AI ethicist and founder of ethical intelligence. Olivia is particularly specialised in how to bring ethical principles in technological development. in order to discuss what are ethics, and how to successfully operationalise them in AI using firm. Going further, Olivia explained implementing ethical principles further will only benefit a firm in the long term, consumers trusting an ethical firm more. She further discussed the requirement to be brave when analysing ethical principles in a firm, overcoming the fear of speaking out against certain unethical practices being the key to make sure the firm is complying to ethical regulations! We thank Olivia for this very insightful talk and detail her points further here:
We talk about creating an ethical AI, but what are ethical principles?
Olivia loves saying she is a philosopher, more specifically an ethicist. She defined ethics for us, explaining they help us understand morality, hence what is right or wrong. Ethicists research how to apply high level abstract principles to concrete situations using logic, to determine what behaviours/actions are right or wrong. For example, an ethicist would study how privacy is taken into account in a new software that uses consumer data and AI. By applying the abstract ‘privacy’ principles, the ethicist could conclude whether the technology abides to privacy rules or not, therefore whether it is a ‘good’ technology! While today, ethics has largely become a buzzword used for marketing purposes in firms, she tells us it is vital employees understand what ethical principles are and how they could be enforced in their firm. Olivia cited many abstract principles that could be taken into account by a firm for it to be its best possible self, such as trust, wellbeing, privacy, transparency. Naturally, it seems logical a firm should encourage trust and wellbeing, respect privacy and transparency, but the weight the firm gives to each ethical principle will depend on the product developed, and justifies the need of an ethicist capable of hierarchy principles in light of the technology used. This rating of ethical principle’s importance however renders ethicists work subjective, as he decides what is the ‘wrongest’ thing to do. Olivia explains the idea that ethics are subjective is the main argument against implementing ethical regulations in firms, critics arguing since ethicality is subjective, it will not objectively notice what is good or bad in a firm, the regulations not produce satisfactory improvements within a firm. Mostly, individuals going against ethical standards of a firm often fear it will reduce their innovation capabilities. Going against both of these statements, Olivia explains how ethics are firstly as subjective as any other science, and furthermore that implementing regulations drives innovation more so than limiting it!
Ethics are subjective, but so are data-sciences!
Questioning here the claim that science is objective, Olivia explained ethics are really no more subjective than data science! indeed, according to Olivia, objecticity just does not exist, ‘truth claims’ always being mediated through the lens of the individual making the claim. As she explains, sciences known to be ‘objective’ such as data science have been in fact subjectively produced by data scientists! As we see in AI biases, the scientist can initially feed incomplete datasets in the algorithm, which creates flawed, biased results. Despite machine learning processes where the technology is said to be independent from human interventions, the initial data provided can through machine learning processes create biases different to those that would’ve been created by humans, but biases nonetheless. A good example of this would be when algorithms are used to rank CVs in admission processes. This method, usually used to provide objectivity by limiting humans’ discriminatory interventions and ensuring CVs are not ranked based on race, gender or disability can sometimes backfire. Indeed, in 2018, Amazon famously ditched its AI recruiting tool for it realised that it was consistently over recruiting men. Indeed, the data they had initially fed their algorithms, which consisted of past successful and unsuccessful CVs, was biasing the recruitment tool. Indeed, since for years recruitment teams had preferred to employ men and had purposefully discriminated women, the algorithms, without understanding what men or women were, had started to consistently associate the word ‘women’ in CVs with a bad resume. Indeed, while the CVs did not disclose gender, some activities, such as ‘women’s football team’, or ‘women’s book club’, had been noticed, and systematically associated to a lower grade CV by the algorithm. Hence, while the machine was objective and could not discriminate based on gender, it ended up doing so because it had been fed biased data! The recruitment policy had backfired. This simple example just shows how ethics ‘subjectivity’ is not a valid argument for its non-implementation within a firm, and should not hamper the importance of ethics within a firm.
“Olivia explains how
ethics are firstly as
subjective as any
other science and
innovation more so
than limiting it!”
Why ethics should be widely used and encouraged in technological development
Olivia then further turned to why AI must be needed in technology. More so than a choice, Olivia calls for a need for firms to embed ethical principles in their technology development in order to thrive. She argues ethics are a risk mitigator, and an innovation stimulator for firms. Indeed, implementing transparency, equity or wellbeing principles in the firm or technology developed reduces possible scandals later on which could hamper the firm’s reputation. Of course, many regulations oblige AI firms to adopt ethical principles, but adhering to these principles whether regulations exist or not will always help reduce future negative risks linked to the technology’s deployment.
More so than a risk mitigator, Olivia argues ethical principles can promote innovation within firms. Ironically, firms benefit from being constrained by ethical principles, as the added layer of regulation they may face encourages them to innovate creatively to go around the obstacle, hence they produce, in the end, a better product. The firms can also better understand what is needed of them, and the way the ideal product should be. By being limited by both technical and ethical limitations, the firm is framed. Overall, good ethics equals good business in the firm on the long-term, as it drives innovation and limits future risks for the firm, ethicists also noting that consumers will likely prefer and trust a product if they feel it respects privacy and transparency principles, the product or firm being viewed as doing something ‘good’. Good ethics within a firm thus encourage long-term stability and rentability.
Operationalising AI- how ethical methods can concretely be implemented in firms
Finally, Olivia ended by concretely explaining how ethical principles could be operationalised in firms. While regulations that ensure some ethical principles are respected exists, such as the GDPR, Olivia explained the best ways to deliver these regulations within the firm. She noted 2 main methods/approaches:
- a top down approach, more suitable to larger firms, that would concentrate on establishing ethical frameworks through a centralised ethics department, able to encourage ethical practices and actively monitor and manage their implementation. While this method is very process heavy, it is very beneficial for large firms who can benefit from a large centralised ethics department capable of filtering ethical principles and communicating new regulations.
- A bottom up approach, more suitable to small firms, that focus on training individuals on ethical principles rather than obliging divisions to comply with regulations. This method works well in small firms, as with limited employees very prone to work together, all individuals in start-ups must be aware and educated on important ethical principles the firm’s technology must abide by. This encourages firm members to raise ethical issues within the firm.
The importance of bravery
To conclude this very insightful talk, Olivia presented one of the most fundamental characteristics of an AI ethicist: bravery. Sometimes, it may seem daunting when one is employed in a firm to go against the status quo and denounce unethical practices, whether it be employment discrimination, wronged use of data, or little transparency of the production chain. However, in order to really change things, it is up to both consumers and employees to stand up to their ethical principles and make their voice heard, in order to promote long-lasting, ethical transformations within the firm!
Olivia provided a reflection on why ethical principles matter at the firm level, and how, contrary to popular belief, rather than limiting a firm’s innovation capacity, they reinforce knowledge development and innovation in the long term. She further explained how to concretely operationalise AI ethics in different firm settings. We thank Olivia Gamberland for this very insightful talk, and hope to hear from her soon!
Written by Jeanne Rouot