I wrote a reply to this earlier but I’ve just noticed it doesn’t seem to have gone through.
To sum it up – I currently fly the 737 for work.
I saw this video recently and having a few weeks ago flown an approach in very gusty conditions I asked our safety team for a FDM (flight data monitoring) replay of the approach. This shows all the flight parameters and control inputs etc. I noticed that when the autopilot was in the control inputs were fairly minimal and not too frequent. Whenever the autopilot was disconnected the control inputs (my own) became more substantial and frequent. I would not agree that I was over controlling however as I distinctly remember that the reason I disconnected the autopilot at that time was because my colleague and I agreed it was not performing satisfactorily in the conditions and it was not making a very stable flight. Once doing it myself, I found that despite me using more frequent and greater control inputs, the result was a much more stable and comfortable approach.
I believe that it is entirely possible to go too far and end up in a self induced oscillation through over controlling but I also would say that certainly in the 737’s that I fly sometimes it just requires a bit of positive interaction. If I had on that day have disconnected the auto pilot and then not made corrections to the distubences, but instead let it ride it out waiting for the opposite disturbance, it would have been very unstable, very uncomfortable and very disconcerting and would almost certainly have led to a go around under our company procedures.
Maybe it’s a Boeing thing 
Pirho wrote:
I distinctly remember that the reason I disconnected the autopilot at that time was because my colleague and I agreed it was not performing satisfactorily in the conditions and it was not making a very stable flight.
You just destroyed the image I had of big iron autopilots and all of the sudden the KAP140 may not be so bad afterall 
Not long ago our autopilot was flying down the ILS. It looked very much like in that video. So I asked my colleague (who was pilot flying) if he was feeling confident to take over manually because I was expecting the autopilot to disconnect itself any moment. So he disconnected the AP and guess what? The last 1000ft down to the runway were smooth as silk with the control columbo hardly moving at all.
what_next wrote:
Turbulence is a stochastic process. For every gust from the left there will be one from the right which cancels it out.
Chop will be, but other kinds of turbulence (thermals, orographic) won’t. For instance if your right wing passes through a thermal that your left wing doesn’t you will roll to the left but there might not be a compensating gust from the other side of the plane any time soon. Similarly, curlover from hills, wake turbulence, rotor etc.
Thermals, rotors and so on are not turbulence in a strict definition. Turbulence is per def a stochastic process.
what_next, that is quite interesting that it seems almost the EXACT opposite of my own experience on that particular occasion I referred to. I tend to find the 737 A/P quite lazy in rough conditions and you can often witness it achieving a phugoid type path around the desired one (both vertically and laterally) which can be maintained with relative ease during manual flight, sometimes doing almost nothing and riding out gusts works, but as orographic effects and structures influence things (as alioth points out) I find more input is needed.
LeSving wrote:
Turbulence is per def a stochastic process.
In an aviation context “turbulence” does not necessarily mean “turbulent flow”.
Well, what people mean and say can be different things. A glider pilot would never describe a thermal as “turbulence”. A rotor is not “turbulence”. A vortex is not “turbulence”. A wing tip vortex is often used as a school example of nearly perfect potential flow for instance, as far away from turbulence as it is possible to get.
If turbulence is “shaking” in any sort of manner, then yes, turbulence it is. A gun vs rifle debate
IMO turbulence imply randomness. If it’s not random, then it’s not turbulence and can be predicted. The predictability is a key factor here when you want to control an airplane.
LeSving wrote:
A glider pilot would never describe a thermal as “turbulence”.
But a power pilot would, and a B737 is definitely a powered aircraft.
In powered flight, air that results in anything other than smooth flying is called turbulence. It’s even in some of the names, e.g. “wake turbulence” which is a vortex. Turbulence near the ground where aircraft are making instrument approaches is often orographic in nature on a windy day and you’ll often not get a bump that causes a left roll to be promptly cancelled by one causing an equal right roll.
