I recently attended a Venture Beat webinar entitled: 3 Keys To Moving Toward White-Box, Explainable AI.
The panel discussed varying degrees of transparency in respect to understanding the output from artificial intelligence, the worst being a black box where the underlying decision is masked in AI algorithms.
So what does this all mean?
Of course, in the travel industry, we traditionally have a different term for a “black box”, but in the context of AI, a black box output implies that it may be challenging for the target audience to understand the rational for a given decision.
The key here is defining the target audience.
For example, data scientists may well understand why a specific output was produced by machine learning, but what about the other business stakeholders?
Indeed, how are AI decisions being explained from a C-Level perspective?
Clearly in defining decisions driven by AI, the audience varies in degrees of technical understanding and focus.
The webinar went on describe a spectrum of degrees of explainability from a black box to a white box with varying degrees of gray in between. The key is to know your audience.
How can black, gray or white box AI impact the travel industry?
This question depends on what is being recommended and to whom.
If an airline is using sophisticated AI to optimize their schedules, it is unlikely management would need to fully understand the parameters of the AI decision - the key measurement is on how the resulting schedule optimizes equipment, people and profits.
On the other hand, if a passenger interaction is involved, the need for a white or gray box solution may be more essential.
Most would agree that the most opportune time to provide some soft compensation (e.g. a seat or room upgrade, additional loyalty points, etc.) should be as close to the disservice as possible.
As we move to a more personalized travel experience, airlines, lodging and car rental companies will need to operationalize their AI to deliver the right service to the right passenger at the right time.
These clichés are fine, until you actually start implementing additional services at the point of traveler experience.
If I am upgraded to a better cabin or room based on a recent poor experience, a more loyal customer may observe this activity and feel cheated.
This brings to light an important aspect of implementing AI, if the traveler is directly impacted by the AI decision, the travel industry needs to be as transparent as possible (gray to white box) in explaining the reason for the decision so the traveler can understand why the decision was made and why they received or did not receive a given benefit.
Without this eye to transparency, as AI becomes more mainstream in the industry’s decision-making process, unexplainable AI decisions could ultimately cause additional frustration for travelers.