Transparency and Explainability – Two key principles of AI 

In the field of Artificial Intelligence (AI), two fundamental principles play a crucial role in building user trust: Transparency and Explainability. These principles not only enhance our understanding of AI-driven decisions but also contribute to the continuous development and refinement of advanced technologies like artificial intelligence.

Among the nine key principles proposed by Humane-AI, Transparency and Explainability are the first two, serving as the foundation for ethical AI development and application. These principles are essential in ensuring a sustainable digital society—where cutting-edge technologies coexist and directly impact human life.

What is Transparency in AI?

Transparency in AI refers to the principle that AI systems should provide clear, understandable information about their operational mechanisms and decision-making processes. AI—particularly deep learning models—is often compared to a “black box” due to its complex and difficult-to-interpret internal workings. This lack of clarity raises concerns, as it becomes challenging to determine why an AI system made a particular decision, especially when errors occur. When AI is transparent, its processes can be monitored, scrutinized, and better understood. This fosters trust, ensures fairness, and allows human oversight when necessary.

A notable case highlighting the issue of AI transparency is the Apple Card controversy in 2019. Apple Card, a credit card issued by Apple in collaboration with Goldman Sachs, faced allegations of gender discrimination in credit limits. Numerous customers reported that women were receiving significantly lower credit limits than men, despite having comparable financial profiles. This led to accusations that Apple Card’s credit assessment algorithm was biased against women.

This incident underscores the importance of transparency in AI systems, especially in sensitive sectors like finance, where AI-driven decisions can have significant consequences for both individuals and businesses.

What is Explainability in AI?

Explainability refers to an AI system’s ability to provide detailed insights into its decision-making process. An AI model with strong explainability allows users to understand the reasoning behind a specific decision, rather than simply receiving an outcome without any justification.

Returning to the Apple Card incident, software developer and entrepreneur David Heinemeier Hansson publicly shared on social media that his wife was granted a credit limit only 1/20th of his, despite their shared assets and joint tax filings. This sparked widespread attention and led to an investigation by the New York Department of Financial Services into allegations of gender bias in Apple Card’s credit limit assignments.

Had the Apple Card AI system been able to explain why it assigned significantly lower credit limits to women despite comparable financial backgrounds, it would have been considered an AI system with high explainability.

The benefits of adhering to Transparency and Explainability principles

Following the first two principles outlined by Humane-AI offers several advantages for both AI developers and users:

• Traceability: Transparency requires that datasets and decision-making processes be documented and traceable. This ensures accountability and allows users to verify information when needed.
• Understandability: Explainability helps both developers and users comprehend AI systems, reducing concerns about accuracy and reliability.
• Communication clarity: AI systems should effectively communicate their functionalities and limitations to users and stakeholders. This enables users to leverage AI appropriately in work and daily life.

For FSOFT, adhering to the principles of Transparency and Explainability is a critical step in ensuring that its AI systems are not only functional but also trustworthy. Providing transparent AI systems that can justify their decisions creates a safe, accessible, and responsible technological environment.

Today, in key industries such as finance, healthcare, and education, ensuring AI transparency and explainability can help mitigate risks and enhance trust. This, in turn, paves the way for broader AI adoption across various sectors, driving positive societal impact.