The whole "wave-particle duality" mess is just that: a misleading mess, and it better if one banishes this term from one's repertoire altogether.
First off, what we observe that leads to positing this notion is this:
- At very low intensities, when light hits a detector, it hits it like little speckles. This is perhaps the most direct evidence suggesting that light can be thought of as being composed of particles. There are actually some more pieces we need, some of which are more complex, to make it airtight, but it's a good start.
- Moreover, we find that the energy of light at a given frequency, likewise, accumulates in anything that is put to do so, in regular steps or jumps, the size of each jump being proportional to the light's frequency, again, as though it were composed of little bits with a set amount of energy in each one. This was what led Einstein to postulate the photon.
- Bulk light - i.e. where the particles are too numerous to count, when we shine it through an opening that is suitably small, or we try to view things that are suitably small under a light microscope - shows effects that look like wave propagation: we get interference patterns and diffraction blurring of tiny objects (which is why that electron microscopes were developed, and that further shows and was actually based on, that similarly, bulk electron beams also act as a wave).
The first two points favor the "particle" concept, while the last one favors the "wave" concept. Historically, the last point of evidence came before the first two, as the first two require considerably more technological sophistication. We may ask, then, as to how we can reconcile these notions. One way to do that would be to take such a bulk wave setup and then diminish the intensity of the light to the point where the effects in the first bullet point become noticeable. When we do that, we observe this:
- The light hits the detector behind the opening like little speckles, as before. But if we record their positions and let them build up over time, the speckles begin to trace out the wave pattern. Moreover, this happens no matter how few we let through - even one at a time.
So what does this mean? Well, we also have one more important piece to take into account:
- Under no circumstance can we ever register a single "particle" of light as though it were some kind of physically-extended object, like a wave.
To me, it is these last two points that put the kosh on this notion. We can, of course, posit that perhaps maybe it's a wave when we're not looking, and always is when we are, but that is about as valid or not valid as positing that there is always a silent gnome behind your back when you aren't looking, and never there when you turn around to look. It's possible, but then again you can also insert an infinite number of other things.
Instead, the most cogent idea is that (we should think) it is a particle, but the propagation of such particle from emitter to receiver is (highly) nonclassical, and in "just so" a way that each such particle ends up having the capability to single-handedly contribute its own part to the wave pattern that would be formed if one sent a "true" wave through the two slits. Getting into the details of how our best descriptions of this work is a different matter, saved for a different post. To answer the question as posed though more head-on, "wave-particle duality", if you want to call it that, means that particles have the ability to create seemingly non-local and/or holistic patterns even without all the pieces of the whole being present at the same time, so long as they all come together eventually.
Ultimately, science does not give "absolute truths": rather it gives useful stories we can tell about the world. It doesn't tell us "what things are", but rather provides ideas we can use to reliably answer questions about what will happen if we do/don't do something, or what we will see given a set of circumstances, and it is up to us to figure out which ones maximize their utility while sacrificing as little understandability as possible. Scientific "truths" are really about actions and consequences, not about things. Scientific models, or theories, are social constructs. What aren't, are the guidance they provide.