Ethical AI: Key Considerations for the Manufacturing Sector

Although there are still no clear answers and the debate continues to evolve, these fundamental questions are more important than ever and manufacturers — like everyone else across the business spectrum — will have to address them as they strive to respond to digital acceleration.

The importance of ethical AI

Machine learning (ML) and AI technologies have been prevalent in industries such as finance and advertising for years, but for manufacturers, the adoption journey has only just begun.

While some would say that manufacturing has fallen behind, the truth is that now is the perfect time for manufacturers to commit to AI investments. Their peers from other industries didn’t have, at the time of their investments, the mature understanding of AI and its ethical impact that is accessible today.

As such, manufacturers have the opportunity to start on the right foot and adopt AI responsibly. Their value-based consumers and employees will highly appreciate it, especially because, for all its effectiveness and decision-making prowess, AI is still just an advanced tool for computation and analysis that is susceptible to errors and even bias if trained or fed by the wrong hands.

Ultimately, this tool can be weaponized to destroy the very same areas of life and business that it once augmented, including public safety and cybersecurity. And that’s why ethical AI is so important, particularly in a world that is becoming increasingly digital at an unprecedented pace.

During the 2020 Aftermarket Virtual Summit, Christian Baudis — Digital Entrepreneur and Former MD Google Germany — hosted a keynote session titled “Manage Your Company’s Digital Strategy” in which he wisely remarked that digitalization will reshape the future of society and that using data the right way will be (and already is) a competitive advantage.

In the future he described, humans will ask the supercomputer what to do at any given chance and, in such a future, having high-security standards is mandatory. And this can only be made possible by having strong legislation in place.

Until further notice

Undoubtedly, AI requires a new policy. However, planning and enforcing effective technology policy is no easy feat. Until governments, associations, and industry circles across the globe reach a more profound consensus on the matter, the 2019 EU ethics guidelines for trustworthy AI outline seven key requirements:

• Human agency and oversight;
• Technical robustness and safety;
• Privacy and data governance;
• Transparency;
• Diversity, non-discrimination, and fairness;
• Societal and environmental well-being;
• Accountability.

Although many dismiss these guidelines as ill-defined and frown at their lack of specificity, it must be noted that they were very likely not created by technical personnel with practical experience in manufacturing. As such, it would be advisable to partner up with authorized consultants or providers that have already figured out how to build ethical AI frameworks for specific sectors.

Key considerations for manufacturers

If you are serious about ethical AI, you should prepare yourself for an intense journey that will take you across different departments, including legal, data science, and risk management. Couple that with outside expertise and you’re on your way to making real progress in applying the principles of ethical AI to real-world deployment environments. 

The recommendations below will help you get started on the right foot and prepare for the long haul.

1. Establish clear metrics

Your journey to ethical AI should start with a collaborative effort to establish clear metrics that can be monitored and measured by engineers, legal experts, and data scientists. So, put together a team of legal and data science experts to help you operationalize data and AI ethics by translating seemingly vague principles into concrete metrics. 

This will help you understand when and why ethical failures may occur within your AI system. Additionally, it will also help you quantify the damage created by AI which will enable you to take effective mitigation and prevention steps.

2. Remove biases

Humans are not the only source of AI biases — data can also contain biases that your AI system will feed upon. Unlike human intelligence, AI is not capable (yet) of autonomous, morally conscious rectification, if it is not explicitly taught that it made mistakes and the data it fed upon was harmful. 

However, you can ensure as much fairness as possible by thoroughly investigating the decision-making process and enforcing standards to make sure that any human bias — which is often unintentional and unconscious — is eradicated. This will reduce the risk of AI misinterpreting biases and proliferating errors.

3. Accept accountability

No matter how many experts you manage to bring together and no matter how error-proof your ethical AI strategy might sound in theory, the truth is that AI is a human-developed tool and the ones handling it are humans as well. Human bias may creep in at one point or another and accidents — such as self-driving car crashes — are likely to happen.

Understand and accept that AI presents liability risks that you need to be prepared for. Negative AI outcomes can impact your entire business and everyone involved with it, including employees, vendors, customers, etc. Make sure that you have a strategy in place to prevent and deal with such scenarios, just in case.

Look on the bright side

Legal precedents, existing research, expert advice, and ethical AI best practices can help you successfully adopt AI in a responsible manner that serves both your business and consumers.

To leverage the multitude of benefits AI can offer — including the automation of manual labor, reduced production costs, and increased organizational value — you must commit to making this technology and your use of it transparent and fair.