EthicalAI is a dimension of responsibleAI that is mostly concerned with fairness. In other words, that AI treats end users, customers or consumers equally. It’s gaining more attention because of recent incidences of output bias such as:
- Google’s settlement of $118m in an ad targeting discrimination lawsuit
- Facebook’s algorithm that showed gender biases against classified groups when promoting ads.
- Uber has been receiving criticisms of racial bias for using ML algorithms to determine fares based on the suburb status of riders. Uber drivers have also accused the company of racist facial recognition software.
- Discovery Health has been accused of algorithmic bias against black healthcare professionals when investigating fraud, waste and abuse.
- Apple’s algorithm to calculate the credit score of customers applying for credit limit increases, was found to be gender-biased, favouring males over females.
Other dimensions such as data privacy and security have been more top-of-mind as most companies are moving their customer data to the cloud and millions of people are sharing their personal data on cloud-based platforms. Think of all those apps we sign up for.
Although fairness breaches don’t sound as scary as major data leaks, companies face potential reputational damage if people believe they are being discriminated against. and material damage if products or systems need to be recalled. See for example:
- Nikon S630 model was biased against Asian ethnicities, when capturing facial images
- Hewlett-Packard’s MediaSmart webcam was flagged because it showed that its camera could track only White user’s faces not Black faces.
So, what’s my interest in Ethical AI? I’m not a computer or data scientist but I am very intrigued about the impact of AI, specifically Gen AI, on our psyches. I love Gen AI don’t get me wrong. I use it for research, analysis and for graphic design. But the full effects of how people engage with it is not clear and there have been some devastating consequences. Which is why I’m studying cyberpsychology. I believe in the beneficial potential of AI but I think it can’t be developed without the oversight of social scientists, cognitive psychologists and, yes, cyberpsychologists.
It means that when Gen AI perpetuates stereotypes or generates outputs that discriminate against certain groups of people, it’s going to worry me. Most significantly the impact it has on the psychology of people who are impacted.
It’s nice to see big tech companies talking about ethical AI and some have developed programs to detect bias in machine learning processes for example:
There are also a multitude of visualization and explainability tools or APIs that can show discrepancies between groups or embedded bias in text.
It’s important therefore that companies investigate the AI models they are using for potential bias and it’s better to start at the very beginning ie. the problem formation phase.
