The Challenge: Overcoming Bias & Misuse
We are currently in the fourth industrial revolution: digital transformation. The exponential growth of data, combined with the rise of AI, is rapidly reshaping our lives. The way we think about and use data; the way we engage and communicate; the way we make decisions, operate devices, experience daily activities — it’s all impacted by digital transformation and the increasing prevalence of AI.
The Solution: Core Values and The Six Pillars
As a premier AI & data company, Lixar understands that it is critical to set standards that ensure our approach to AI is in line with our core values: 1) enable people; 2) design for inclusivity; and 3) build trust in technology. As a Microsoft partner, we have built off of our core values and Microsoft’s AI principles, to establish the following six pillars for maintaining strong ethics in our AI practices.
The Six Pillars:
- Fairness: AI systems should treat all people fairly and without known bias
- Inclusivity: AI systems should engage everyone and be usable across the ability spectrum
- Reliability & Safety: AI systems should be reliable and safe to use
- Privacy & Security: AI systems should be secure and respect privacy
- Transparency: AI systems should be clear and concise in their intent and function
- Accountability: AI systems should have algorithmic accountability
AI systems should treat all people fairly and without known bias. To maintain this pillar, it is critical to avoid knowingly training AI using biased data, as any such bias could be reflected in recommendations the AI would make. For example, on a job recommendation platform, men and women with the same professional credentials should be recommended for the same jobs, with equal compensation. Upholding this pillar will ensure that people are treated as equals and are given access to equal opportunities. It is key to enabling people.
AI systems should engage everyone, and be accessible to users across the ability spectrum. Maintaining this pillar means that we have to consider diverse perspectives and experiences when developing AI. Similar to fairness, a known bias towards one kind of user’s lived experience will make for a poor user experience for many others. A chatbot AI, for example, that can only communicate via text and in one language, will not be engaging nor accessible to people who are vision impaired, have difficulty reading or speak another language. Designing for inclusivity will ensure that the technology empowers as many people as possible.
Reliability & Safety
AI systems should be reliable and safe to use. Upholding this pillar means building AI with the highest quality possible, every time. Delivering products that are unreliable or unsafe would be strictly against our core values to enable people and to build trust in technology.
Privacy & Security
AI systems should be secure and respect privacy. To uphold this pillar, it is critical to provide expert quality assurance and testing, as being powered by data means relying on professional-level storage and maintenance. In addition to having strong QA and testing protocols in place for all our products, we are knowledgeable and capable of providing services that include planning for disaster recovery and restores. As with Reliability & Safety, this pillar is also key for enabling people and building trust.
AI systems should be clear and concise in their intent and function. To maintain this pillar, we must be dedicated to integrity in our work, supporting an open dialogue about new technologies, their purposes and any subsequent results of their use. Moreover, when the technology is easy to understand, it’s easy to use and can be more effective overall. For example, an AI prediction will have more of an impact if it’s clear how and why it came to that conclusion. Something like data visualization can go a long way for transparency. Promoting this pillar will help address people’s concerns about the technology so they can comfortably embrace its benefits.
AI systems should have algorithmic accountability. Maintaining this pillar requires being open about the purpose, structure and underlying actions of the algorithms used to search for, process and deliver information. Going hand-in-hand with transparency, accountability allows us to identify vulnerabilities in the system and take the steps necessary to correct them. We believe strongly in keeping humans in the loop; in the event that an AI makes an undesirable or incorrect decision, it is key that a human be able to immediately override that call. By upholding this pillar, we can continue to progress the technology, ensuring that it is respectful and always to the benefit of people.
End Result: Ethical AI
AI should be used to empower people and to contribute to the overall good of society. Being ethical in our approach to AI is all about developing systems that are fair, inclusive, reliable, secure, transparent and accountable. When we do this, we create AI that has the potential to make the world a better place.
To learn more about AI and other data-driven technologies, drop us a line at firstname.lastname@example.org.