RWA:
Based only on my understanding of the resource you cited, and a slightly dated training manual on the A330, the automatic feature (alpha protection) would induce control inputs (you could say "fight the pilot" or "help the pilot since what he's doing is heading in the wrong trend") in normal law, but in alternate law, that AoA related feature doesn't (should not?) do that. Stall warning should sound if the AoA is deciphered as a value approaching stall. (It appears that it did ... but also that there are some quirks at very low airspeeds regarding that feature).
What is frustrating in understanding this incident, for me, is the number of parameters that need both Airspeed and AoA as inputs, and the coupling of Airspeed to AoA, and problems this poses for the flight control system when those fundamental inputs are spurious or obviously erroneous, or "voted out." (Note: since the mishap was reported, there has been a Greek Chorus of pilots who point out that flying pitch and power is where one begins to deal with erroneous airspeed problems ... and there is also a considerable body of pilot experience that points to hand flying at altitude becoming a lost art ... )
Systems complexity considered, once in alternate law the speed protection can be overridden by pilot. Whether or not the computer sensed an overspeed protection requirement (and thus a kick off of the autopilot) has been discussed at some length. The consensus seems to be that AP kicked off due to crap airspeed inputs, not due to overspeed sensed (erroneously or otherwise).
As the bits and pieces of info come out of the investigating team, the discussion on Tech Log forum has been skeptical of initial (via the high speed protection feature?) NU input commanded by the robot. Part of the support for that skepticism has been the kinetic energy tradeoff (q) and the recorded pilots back stick inputs (along with roll inputs). These inputs look to have accounted for the pitch attitude and altitude gain (and of course, actual airspeed loss, versus displayed which was for a while all wrong).
Originally Posted by sources albeit dated
High speed protection is active, depending on flight conditions, at / above Vmo / Mmo, and a pitch up load factor up to +1.75g is added to the pilot input when speed exceeds Vmo+6kt or Mmo+.01.
Speed is limited to Vmo+15kt or Mmo+.04 even with full forward stick.
If full forward stick is applied suddenly, speed is limited to Vmo+35 or Mmo+.07 and then returns to the above values.
The protection is deactivated when the speed returns to V/Mmo.
The AP will disengage if the high-speed protection is active.
In Alt 1 ... Above V/Mmo an over-rideable nose up demand is introduced. ... Alternate 2 Pitch control laws are identical to Alternate 1
Protections are as in Alternate 1
... in the case of failure of 3 ADRs, no high speed protection.
What I glean from this is:
if all three of the sensors that feed airspeed info to the flight computers and displays went pear shaped -- the evidence points to this based on pitot icing -- then the automatic pitch up would NOT happen since in Alt 2 that feature is disabled. This presumes that the flight computers more or less were functioning as advertised.
Whether or not a suprious voltage or oddball computer decision entered into this cannot be known as of today. It may be unknowable, if such transpired. There does not seem to be much of any evidence to support that class of malfunction. If such did happen, it didn't leave much of an audit trail.
Svarin has made some thought provoking posts and analysis in that regard, recommend taking a look at his inputs if you have not already.
Cheers