Page 11 - summer of ai

SUMMER OF AI JULY 2019 DATA & ANALYTICS
Opinion
Ethics Behind AI in the UK
In the absence of regulations for artificial intelligence, conversations about AI in Europe center on ethics
D“Take care of the pence, and the pounds will look after themselves.” ating back as early as 1747
in a letter about the British Secretary of Treasury, the proverb means if people try to save small stuff (like pennies
or pence), you will naturally save on the big stuff (like dollars or pounds). Parliaments across Europe are thinking about the ways in which artificial intelligence will help consumers obtain prosperity by ensuring a fair credit decision.
While the British government has yet to regulate artificial intelligence, they have taken note of the ethical implications of the opaque decision-making of neural networks. One common emotion I sense in my conversations with regulators across Europe is fear. A real fear that is rooted in the belief that says the only way you can use AI is for it to be a black box. Meaning, an algorithm will determine an outcome that no one can explain. I have one single message to calm their fear: Yes you can illuminate the reasons for a decision if you constrain models as we have done with NeuroDecision® Technology.
You can build an explainable AI. It’s both very clearly explainable and actionable to the extent that you use it to process a loan application
and decide whether or not to grant credit to John Power. (Oh by the way, you can also explain in very clear terms the reasons why you declined John Power too.) Because you’re able to explain the credit decision very clearly and the consumer easily understands, the reasons also become actionable because John can look at the explanations and say: “Okay, I’ve just been told I didn’t get the mortgage I applied for because I’m very heavily utilized on my credit cards because I maxed out my credit cards and by the way I missed
a couple of repayments on my auto loan in the last 12 months.” So, when the consumer receives very clear explanations as to here’s why you’ve been declined, consumers can act on the knowledge. Armed with more information, consumers around the world can now make smarter decisions to live their financial best.
If lenders can use explainable AI to build models that help consumers understand how to take care of the pence, as the proverb goes,
the pounds will take care of themselves. g
JOHN POWER
VP, STRATEGIC PARTNERSHIP AND CLIENT ENGAGEMENT
John Power and his colleagues will present four white papers about different aspects of explainable AI and its impact on the lending industry
• Predictive Models With Explanatory Concepts
• Approximating An Optimal Path To Improve A Consumer’s Credit Score
• Direct Estimation Of The Survival Function For Credit Default Using Nested Time Intervals
• Best Practices In Building Neurodecision® Technology Models In The United Kingdom
11
Visit Credit Scoring and Credit Control XVI Conference for more information.