As customers and regulators increasingly demand explanations regarding algorithmic decisions, having a trustworthy and transparent AI framework in place is a must for any business working with such systems. Especially in the financial sector, where the stakes are high and time is short, a reliable and responsible AI solution can improve efficiency and upgrade quality. Time for a quick Q&A about VIQTOR DAVIS’ FACT AI by-design.
My company already has a rule-based framework in place that automatically handles half of our credit applications. Why would I need a new one?
Half of your credit applications you say? For a major Dutch financial institution, we’ve managed to bump this number up to 90 percent using AI! On top of that, this AI solution is also 12 percent more accurate. Some of the most challenging applications they receive are now handled automatically, saving time and money.
Sure, but then your AI process must be so complicated that no one understands it. I still need to grasp why an algorithm makes decisions and explain it to regulators and clients.
We acknowledge that you need AI that is understandable and explainable in order to abide by the rules and gain the trust of your customers. That’s why we have developed a state-of-the-art approach that implements Fairness, Accountability, Confidentiality, and Transparency in any AI solution – something we call FACT AI.
How does that work?
This is where it gets interesting. And also, slightly technical.
Preferably your AI system is so advanced that it can automatically handle the most difficult credit applications. But there is a trade-off: the more advanced and complicated your AI solution is, the more you run the risk of something called black box decision-making. In short, this is a system where the input leads to an output, but where there is no insight in how the system came to this output. The internal workings remain a mystery.
Our FACT AI approach enables you to open up these black boxes, allowing you to deduce how a decision came to be. You can see the interplay between several key factors like an applicant’s credit score, their field of industry and the influence of any outstanding debts. In the end, our FACT AI solution provides a clear explanation stating how all these factors tipped the balance in favour of or against providing a loan. And we can present it in a nice visual too!
With our FACT AI approach, you can have the best of both worlds: accurate AI decision making and intuitive explanations of individual decisions.
And how do you check whether your FACT AI made the right decision?
For the financial institution we mentioned earlier we organised something we call ‘trust workshops’. In these sessions we checked our AI’s decisions and explanations against those of human domain experts, with great results. The credit experts rated the quality of and usability of our AI’s decisions and explanations for the expert’s day-to-day work with a high average score of 8.5 out of 10! These workshops also help to build trust in our AI solutions as executives and stakeholders gain more insights in how our FACT AI works.
So I can explain decisions to regulators and my customers. What else?
Opening up those black boxes has more benefits. As an executive it is crucial to understand recommendations made by AI in order to make informed business decisions. And if you don’t have a clue of the internal workings of your AI systems you run the risk that those systems will make discriminatory decisions, which could result in legal consequences and a public backlash.
Can I only use your AI solution for credit applications?
Of course not! For whichever goal you wish to implement an Artificial Intelligence solution, we can help you make it fair, accountable, confidential and transparent. Our FACT AI approach enables you to understand your Artificial Intelligence so that you can trust you’re taking informed executive decisions and are meeting regulatory requirements.
Ok, you’ve struck gold. Where can I learn more?