The big news in the driver's license world was the collective annoyance at the fact that New Jersey driver's can't smile in their driver's license photographs. Now there is a solution to this problem, which I'll discuss below but let's go over the basics.
Some jurisdictions use a facial recognition system which can't process strong facial expressions particularly well. It's not just a problem confined to the US, for instance you can't smile in Canadian passport photos either.
Hypothetically the facial recognition system can take a photo and run it in the database of existing photos, to see if the person has applied for/received a license in the past. Facial recognition on this scale is complex and difficult. False positives are major problem of course...if you want to increase reliability, you increase false positives, and it takes time (from a human) to sort through and analyze the system's mistakes. Decreasing false positives means the system may miss out on true copies. The quantity of photographs obviously makes a big difference, running facial recognition on the California driver's license database is quite different from running it on Delaware's database. As far as I know, they don't run facial recognition on all 50 states' database, and it's an open secret that the technology couldn't handle that many photos.
This means the only thing the facial recognition system can catch is people applying for a fraudulent ID card in the same state they already have one. Facial recognition is therefore a complex, not particularly reliable system to detect fraudsters who are too lazy to drive over the state border.
Most states have facial recognition systems that can handle strong expressions (like smiling.) A few states, like New Jersey don't. I'm not sure why they would purchase such a system, the fact that it can't handle smiling implies to me that it's not very good and is probably best avoided. I mean, how much cheaper could the no-smile system have been? (Alternatively, the system which tolerates smiling might even be worse in terms of reliability.)
I can't help but be curious about the company and the people who would make a facial recognition system that can't deal with smiles. Talk about compromises to push a mediocre product out the door. For that matter, I'd be amused to watch a software engineer say to a child "I make software that takes pictures of people that they can't smile in." Not something to be proud of.
If you're trying to get to grips with why the smile is important and why people care, I believe it's because the DMV is a dehumanizing place. You're lining up to get entered into a massive population database, you have to prove your identity (something really basic and fundamental about you) in multiple awkward ways, if a failure occurs you find yourself denied basic requirements of life--such as driving a car. It is a position of enormous vulnerability that humans rarely have to endure (the most analogous experience is airport security.) Some people are quite sensitive to this dehumanization. In fact, I believe that sometimes DMV customer service is actually normal, but because you're coming from such an awkward, dehumanized place, you feel more sensitive to minor customer service failures.
So you try to make the best of it, be as human as possible, smile, and pray that you have a halfway nice picture to reward yourself from all the effort.
Oh and the solution? Just take two pictures. One for the database, and a smiling one for the license. I'm not surprised that several state DMVs (and the system vendor) have missed this simple solution, but it's so painfully obvious that even my worst assumptions about their competence can't explain why they've missed it.