As flight deck automation becomes more reliable - to the point of hardly ever failing - it is becoming more of a human factors problem. The UK Air Accident Investigation Branch makes this clear in its report on the Thomsonfly Boeing 737-300 that stalled and was momentarily out of control during its approach to Bournemouth two years ago.
The AAIB cites a Civil Aviation Authority study "Flight crew reliance on automation", observing: "Pilots familiar with operating older aircraft which had more variable reliability are nearing the end of their careers, and there is a generation of pilots whose only experience is of operating aircraft with highly reliable automated systems." Is the AAIB implying that younger pilots are less good than the older ones when things go wrong? It seems so. Maybe that's because the exercises mandated in recurrent training programmes have scarcely changed since the days of the Super Constellation. So training no longer represents what crews are likely to have to deal with in today's aeroplanes.
The Thomsonfly incident was caused by the crew's failure to notice that the autothrottle had disconnected with the engines at idle, and their late recognition that the airspeed had dropped seriously low. This has similarities with the circumstances of the February Turkish Airlines fatal accident on approach to Schiphol; there, the autothrottle retarded the power levers to idle - uncommanded - but the crew did not notice the power reduction or the speed loss.
To coin a cricketing analogy, crews today are like batsmen practised only in receiving lots of fast, straight deliveries. What they actually need is training for the occasional googly. Like a subtle automation failure, for example.