That's the point!
Don't trust computers.
Even though testing in the airplane industry must be really(!) good - you can only test the cases that you think of before hand.
Now take all the rules of physics, airodynamics, weather dynamics etc. and put well-known and working models together into a silicon-box and let it process all the information from sensors to make the right decisions.
1. well-known ... not yet unjustified (Popper)
2. airodynamics & weather... as far as know models do have problems to predict tomorrows weather... so can we say we understand it or are there possibly conditions we do not even know of?
3. silicon-box ... theoretically it is possible that a single electron gets stuck inside a cpu. Now think of extrem temperatures, condensed water, ...
4. I do not know much about the sensors. Though - current media report- it seems that airbus had recommended to airfrance the exchange of certain speed-sensors...
I work for more than a decade now in the software industry and unfortunately I have seen many projects where everything was flawed from the beginning: the whole planning and design process, the development by underqualified programmers, the testing inexistent and the project management is often a complete joke!
One of the craziest things I saw (from a distance): A project manager had skipped testing just to have a new version of a financial planning software released on time. He did ignore the risk of miscalculating budgets and costs for a large corporation. He would not see the consequenses for company-wide decision-making based on possibly false numbers - as long as he would get his bonus (~500 euros!) for releasing on time.
This guy was lucky - since there were no errors detected. but he put the whole company at stake - and all the others did not know. (That whole company and its' processes were flawed by the way...)
I do not know the software-process of airbus or boeing and I am sure they spend a great deal on testing. And we can certainly, statistically say that their machines work fine (there are very few plane crashes)
The point I am trying to make is that complex-software is very difficult to test - even when it is well designed and well programmed.
And just because of the fact, that you can never really be sure whether your tests are complete or whether you forgot one combination of variables (like wether, aerodynamics, sensor-failure,..) there should always be the possibility to override a computer by the push of one button.
Computers can collect more information and they can react much faster than humans can. But only under specific conditions. Conditions that the engineers could think of. This is true for all technology: your plane, PC, iPhone, cars..
Example from the car industry: lately I brought my German-car to have it serviced. There was a man whose brand new car's electronics had locked his trunk - everything he needed for his next meeting was in there. The key would not work. With some effort the mechanic opened the trunk and explained that the mechanical lock would not open without working electronics.
The owner of the car asked what he should do if the trunk locks up again. The mechanic said in that case they would look after it again... but he could not explain what caused this failure!
I bet that some untested condition e.g. a combination of broken sensor signals, water, temperature, manufacturing problem, material has led to that lock up.
I do not want to buy a car, computer, phone or any device that does not have a manual override. Under regular conditions, computers work well. No doubt to this. But there will always be situations where a human can react better.
Humans AND Computers both have their limitations.
Humans should always have the final say!