Our VP for Data Ethics & Governance, Sophie Chase-Borthwick, was recently part of a panel – the PICCASO Special Interest Group. Sophie joined William Malcolm (Privacy Legal Director at Google), Radha Gohil (Data Ethics Strategy Lead at Shell), and Anne Woodley (Security Specialist at Microsoft) in untangling what data ethics actually means and how best to support it. Here we look at this in more detail.
Often, data ethics is mistakenly reduced into just being about data bias, but ethics in data is much more. Data ethics is defined as “the values, principles, and techniques people/companies can employ to create standards of the right and wrong development and deployment of AI technologies.”
There are many negative consequences for unethical data design but the most significant include:
- Questionable design which could lead to implications such as safety risks.
- Unintended negative consequences, such as individual safety (physical, digital, financial), organizational impact (financial, legal, reputational), or societal (security, stability).
- Misuse of the data.
Establish an ethics-by-design culture
Building off the privacy-by-design approach, ethics-by-design means embedding ethical principles into designing, developing, and delivering your products and services. “We do a Data Protection Impact Assessment (DPIA). I don’t see why we wouldn’t do the same with ethics and ethical assessments” says Sophie. This is especially important for a data strategy service.
Radha agrees: “As AI exponentially grows, we need to deploy learning at scale. AI is growing across teams, not just data scientists. Everyone needs to know the basics so they can interact with AI responsibly.”
William adds “It’s important to own processes and principles. When building our data ethics approach, we considered what matters to us as a company and reflected this in our frameworks.”
A data ethics framework can act as a compass point to show true north when you’re mired in data ethics on the ground. Here are important points to consider:
- Build a framework. Work it out. What does this look like on a global scale?
- Consider what good and ethical looks like for your company.
- Measure against your framework the same you would for any other system. Measure the input and the output against ethics metrics.
Security is important, but don’t try and boil the ocean
One of the most important aspects for ethical data is being able to securely store this data.
According to Sophie, “Security, but from whom? Internal or external actors? Do you use a secure environment which everyone can access but results in privacy issues? We know hackers are reverse engineering algorithms so you should consider who you’re protecting your data from. You could be protecting it from your own data scientists but not the malicious external actor. There’s no one size fits all.”
Anne agrees: “Security can be as complex as you want it to be. Security can also be simple. You should make sure to encrypt data when it’s a function, store it somewhere safe where access is controlled. Use zero trust models and make sure you have visibility across the clouds.”
For Sophie, it’s important “To try not and boil the ocean when it comes to data security. You don’t have to put your whole house in a safe … but you can lock the doors,” she advises.
In the 21st century, data is the new gold rush. Secure your assets appropriately.
Embed data ethics in your procurement
Your company might live and breathe data ethics, but what happens if you tender a vendor who doesn’t? This could undo all your hard work. Nick Graham, Partner at Dentons, who was hosting the panel advises: “Dig down deeper than the sales pitch. Uncover how the model actually works.”
How do you maintain ethical practices when liaising with a vendor? Spoiler: exactly the same way you maintain good practice with any other vendor, for example, security.
As Sophie sums up, “Vendor management isn’t new. You should make sure the vendor has the right checks in place and their data is secure. We know this not just for data ethics but for anything you ever outsource from a supplier.”
If data ethics is the puzzle, data bias is one key piece
Data ethics is the holistic approach, but another crucial aspect of this is data bias. You can read more here in our blog post on how to ‘banish the bias’.
” AI and the Ethical Implications of Bias in Machine Learning (ML) Models”
Data Ethics is a major area for consideration in the world of data, governance, privacy and law. Artificial Intelligence (AI) can perform highly complex problem-solving (such as unravelling intricate cancer diagnoses), but it can also suffer major setbacks (such as the potential for racial discrimination).
AI is outperforming humans at narrowly defined, repetitive tasks, which is the space in which AI excels, there are however some risks associated with AI and during our panel debate, we have invited some leading experts and thought leaders to help us navigate this complex area.