This is a math trick to raise or lower the sensitivity of any variable to changes in another variable on which it depends. In practical use, it is much easier when working with most users, to show them a graph of how service quality would vary with workload, and get them to 'sketch' how they think the relationship works [see the user guide video 32
... but if you want to use this math trick ... take a case where some service quality
(e.g. fraction of service with no errors) was sensitive to workload
(work required / work capacity] then
would mean that if workload = 1.1, service quality = 1/1.1 = 0.91 ... workload = 1.2 service quality = 0.83 etc. [the 'min (1, ..] part is simply to ensure that service quality cannot be better than 1.0, even if workload is less than 1.
service quality = min (1, 1/"workload")
But you may find that service quality deteriorates faster or slower than this, so if
service quality = min (1, 1/"workload" ^ "sensitivity")
... then with "sensitivity" set to, say, 2 and workload = 1.1, service quality = 1/(1.1^2) = 1/1.21 = 0.826, and if workload = 1.2 service quality = 1/(1.2^2) = 0.69.
With a lower value for "sensitivity", service quality is less badly reduced as workload rises - and if "sensitivity" is less than 1, then service quality gets worse less quickly than workload grows ... with "sensitivity" set to, say, 0.5 and workload = 1.2, service quality = 1/(1.2^0.5) = 1/1.09 = 0.913.