Algorithm law, the governance of algorithms and compliance with the law or regulations when creating or using algorithms is going to become more and more important. Algorithms (and data or information) are at the core of many things, like social media, credit scoring systems, automated decision-making systems, robots and autonomous cars. Algorithms and data go hand in hand. Robot software is just a series of algorithms. We need to regulate how people and organisations create and use algorithms.
Algorithm transparency, algorithmic auditing, autonomous algorithms and algorithmic escrow are all terms that you’re going to hear a lot more about in the coming years. In this article, we are not talking about algorithmic regulation – the use of algorithms to regulate conduct or behaviour.
How an algorithm can impact you
- You only read (or are exposed to) content that reinforces your beliefs.
- You are granted a loan to buy the house of your dreams.
- You are notified that you were unsuccessful in applying for a job that you want.
- A billboard changes to display an ad specifically for you as you drive past.
- You are granted a Visa to visit a country within minutes.
The chances are that these outcomes were decided by an algorithm.
So many questions about algorithm governance
- What laws currently regulate the use of algorithms?
- What global universal laws do we need in the future? Requiring algorithm transparency (or algorithmic transparency) is only part of the solution. We must build trust through transparency. We need to explore other forms of regulation to ensure that people can trust algorithms and that their rights are not infringed.
- Do we need a red button that can destroy an algorithm that has got out of control?
- Should algorithms be programmed to report back on what they have been doing?
- How do we ensure the ethical use of algorithms by design?
- Who owns the algorithm?
- When must the owner disclose it?
- When can someone complain to the owner about the effects of the algorithm?
- Should Parliaments be creating laws to ensure that the A is fair?
What is an algorithm?
An algorithm is “a process or set of rules to be followed in calculations or other problem-solving operations, especially by a computer.” It is a lot less mystical than many people think. An interesting thought is that when a computer follows the rules and solves the problem, often an automated decision is taken. Let’s look at some examples. Lots of people are interested in the Google algorithm, Facebook algorithm or Instagram algorithm.
It is not magic.
The Google algorithm
- Is it neutral or does it have a bias?
- Does it favour the dominant opinion?
- Does it give preference to the well-financed?
- Might it be manipulated? For example, by the State.
- Might Google algorithms serve Google’s self-interest?
- Do they threaten democracy by manipulating elections?
The Facebook algorithm
EdgeRank is the name commonly given to the algorithm that Facebook uses to determine what articles should be displayed in a user’s News Feed. Questions:
- Does it create echo chambers?
- Does it threaten freedom of expression and other things?
- People can manipulate the Facebook algorithm to get higher on the list? Can people pay to be higher?
- What if Facebook edits and restricted what appears on your newsfeed? For example, when Facebook bans white nationalism.
What laws currently apply to algorithms?
The USA has introduced a bill in the Senate which aims to regulate algorithms. The bill is entitled the Algorithmic Accountability Act of 2019 and if passed, asks the Federal Trade Commission to create rules which will regulate algorithms and machine learning. It will force big companies to evaluate their algorithms to ensure that they are not turning out unethical and discriminatory results. This bill is likely to be the first of many around the world which will regulate algorithms.
The word algorithm does not appear in data protection law. But an algorithm often results in an automated decision. It is code that decides something. There is a law that deals with automated individual decision-making and this is very relevant to algorithms.
Constitutions or bill of rights could possibly provide solutions for now.
Access to information laws could also help? They might be relevant where it is the public sector who is using the algorithm. And in South Africa, their access to information laws (PAIA) might also be very relevant because it applies to private bodies as well as public bodies. Maybe the grounds for refusal in PAIA can be used to decide when Google must give you access to information. PAIA applies vertically and not just horizontally.
Sometimes, competition law or anti-trust law plays a role.
Some considerations for developing algorithm law
If we need to regulate how people and organisations create and use algorithms, what must we take into account when making law?
- The system must be as free and fair as possible.
- We must increase transparency and accountability.
- Any regulation should be principle-based and not rule-based.
We must strike a balance between the benefits algorithms bring and the possible harm it could cause to an individual.
Some useful resources for further reading on algorithm governance
For further reading see:
- An article called Trust but verify: A guide to algorithms and the law
- The Right Way to Regulate Algorithms
- HBR article on We Need Transparency in Algorithms, But Too Much Can Backfire