Computerized BDO: Would it be more/less effective; more/less offensive?
 

Just curious, would computerizing the BDO process work better from both an effectiveness and a PR standpoint? I have to believe that there's a way to integrate software into a kiosk whereby it can tell, through a combination of tone, shiftiness and changes in radiant heat, whether you are becoming more agitated during the questions (regardless of what the questions are) with the assistance of a camera and heat sensor.

When you step up, you could feed your BP in (would accommodate ones from kiosks, counters and home printers) and based on your responses it would print something (number, symbol, etc.) on there and the TDC would direct you to one of two or three lines (low, medium, high risk). Seems as though it would also be easier to monitor (if something in the software alarmed) by real behavior professionals, and you'd need significantly fewer of them since they're not questioning everyone (maybe one or two per terminal at most).

I'm curious if the current state of AI software would actually allow this to happen effectively, and if speaking to a computer screen vs. what passes for BDO would be more or less offensive?

GREAT - We don't have enough B S ??
"I'm sorry, Dave. I can't do that."

My sense, as someone well-trained in computer science but not specifically in AI, is ... no.

Visual processing is extraordinarily difficult. Facial recognition is starting to come along pretty nicely; there are commercially available software packages that do a rather stunning job of recognizing known faces in unknown pictures, for example. But to take a completely unknown face and be able to analyze it in real-time for signs of deceptive behavior (whatever that means)? I don't see it happening.

I don't think our understanding of the theory behind behavior detection is well-developed enough to encode those ideas as computer algorithms. AI isn't "magic"; someone still has to figure out the instructions for the machine to execute.

Machines would have certain advantages over humans as BDOs: speed, ability to analyze many more factors, lack of fatigue, infinite reproducibility, less susceptibility to bias based on prejudicial factors like race, creed, gender, and so on. On the other hand, machines lack the ability to extend compassion and exercise judgment. (Imagine a stressed-out mother, trying to make it home to see her dying father, for whom she has conflicted feelings, encountering one of these machines and denying her passage because it's picking up on her feelings for her father, not the trip.)

And perhaps the best argument of all is ... frankly, if the technology were capable of such a thing, I think we'd already see it in use. Not at airports, perhaps, but perhaps on a much smaller scale.

But all of that aside ... even if we could do this with a machine, I'm not sure that we should. Whether or not the ability to take a commercial airline flight is a right or a privilege (a debate for another time), I don't think we want that ability to be abridged by a computer which cannot explain its reasoning. "The computer said so" is a terribly unsatisfying answer to give.

I have a decent bit of experience in machine learning, so here goes...

A key part training such a system is having a good mix of positive and negative examples to show it.

As it turns out, we haven't actually had too many living, breathing terrorists lately. TSA's always glad to hold up an example of finding someone with an outstanding warrant, but that's not the same as finding a terrorist. Further, remember that even if you manage to find a real life terrorist (but also remember, you don't know that yet!) standing in front of your scanner, he's likely to be one very cool customer. Lie "detectors" can be gamed. Yours will be as well.

Roughly speculating, you'd have the calm folk (frequent travelers, well-trained terrorists) and the more anxious folk (less frequent travelers, new fliers, people scared of flying, etc.). Good luck choosing what you pick up on.

Not Sure I Understand...
 

Being as the TSA allowed, IIRC, 23 terrorists to fly through US airports and has stopped exactly zero, a blind and deaf chipmunk taking a nap could not do any worse.

This might be an interesting idea, especially if the machine prints out a mark on the boarding pass in a predictable manner and location.

Photoshop would do wonders for those of us with home-printed tickets. And we all know that terrorists can't buy Photoshop to modify their boarding passes... :rolleyes:

We're already jumping through too many hoops and spending too much money to solve a problem that hasn't manifested itself and is extremely unlike to do so in the foreseeable future. We don't need to spend even more money on this.

I welcome neither. When confronted with a coin with two tails -- even if supposedly forged with the image of a head -- and asked to choose, it's a game for gullible losers.

Sorry, no. Everyone, including airline and TSA employees should be subjected to the same screening. X-ray of belongings, walk through / hand held metal detector, Explosive Trace Detection / Explosive Trace Portal. Nothing more, nothing less.

And people like me who are quickly short-tempered and hostile to machines. For example, those horrible call systems that make you say "existing customer" instead of pressing a key. Socially speaking, I see no need whatsoever to be polite to a machine, so I'm not.

Better idea. Dump the idiotic BDO program.