Because the two views – machine or human, are held by humans, then the human is held accountable – us, ourselves, how we think about the situation; the process of thinking.
This is not as philosophical as might first appear; consider the events preceding AF447.
A technical weakness was identified with the probes – a rare form of icing. Solution, modify the probes. Until this could be done, the risk of a multiple failure was mitigated by having at least one modified probe, but if this was not possible, the crew would be trained to manage the outcome of the failure - procedure.
This is a logical engineering (technical), machine view. However, there was a break-out point when considering the crew – not as a weakness which required additional training, but as an asset which could mitigate the threat.
The flaw in the engineering view (manufacturer and regulator) was the failure to realise that the threat was the ice crystal conditions, not the resultant of an icing encounter with deficient probes. The alternative realisation required a human, operational viewpoint.
Thence in true TEM fashion, avoidance should have been the primary human activity, supported by information about the threat and how to detect it. Avoiding Cbs involves a routine and already trained for-behaviour, with procedural backup in the event of mismanagement (error).
Re automation dependency; it’s not the pilots who are automation dependent, but all levels of the industry. Manufacturers and regulators have difficulty thinking outside of their numerical, probabilistic approach to aircraft design and certification, whereas the real world involves the indeterminate behaviour of humans interacting with a machine. If regulators become locked-in by a machine view then blaming the human is a logical outcome.
‘A problem to control, or a solution to harness’.