Perhaps it is worth summarising here what science can do and what it can’t do, because this is not made particularly clear in the article. You might imagine scientists had isolated and purified Evil in the lab. The scientists should be wearing goggles, and the Evil in a test-tube glows green.
Let’s start with something less controversial: colour vision. We can measure the sensitivity of the eyes colour receptors. We can measure the power at each wavelength for light. These are things we can measure objectively. Then we know and name the colours we see. We can tell when two colours match.
There are also very convincing illusions where one colour appears to be another because of the context, or the surround. This rabbit-hole goes very deep - we see the yellows in bananas differently to the yellows in (say) lifejackets because we have different expectations. So, we tend to have simple models such as ‘colour constancy’ where we compensate for different lighting, because that is what our brains and eyes are very good at doing. The ‘colour constancy’ model is not an accurate simulation of what our nervous system does, and it does not attempt to factor in (say) our understanding of bananas - but it does seem to fit the way we see things in enough cases to be useful.
In the same way, we can have a feeling of what ‘intelligent’ means, and we can measure the ability to do IQ tests. If we have done a good job then the IQ score ought to fit our sense of ‘intelligence’ in the person, and nothing else. This is really hard to do, and usually the most we can do is to rank people with the same background.
With the ‘D’ score, they experimenters are asking different types of questions to try and score people on different recognised types of ‘darkness’. Suppose we were to measure (say) three different types of darkness, and plot the results in 3D. If all three forms of ‘darkness’ were an exact measure of a single parameter, our plot would be a diagonal line. With real people, we expect to get some sort of rugby-ball shaped fuzzy blob. We can measure the shape of this blob (look up ‘principal components’ or ‘eigenvectors’ to know more), and we can say to what extent the different forms of ‘darkness’ correspond to a single form of ‘evil’. And that’s just what they are doing. It doesn’t pretend to tell us what ‘evil’ is, but it may give us a model that fits what we call ‘evil’ in enough cases to be useful.
Kinda dull, innit? But science has to start somewhere, I guess.