CycloWizard's comments on the scale of things is a good view. From a chemist's perspective, though, an important factor is the composition of the "particles" dispersed in the "medium".
In a true solution, each solute "particle" is a single molecule, and the "medium" is the dominant component we usually call the solvent. So a simple solution has single solute melocules surrounded by many solvent individual melecules. This does NOT preclude some associations between individual molecules - sometimes a few solute molecules might stick together weakly, but they can be broken apart by small disruptions. Or very commonly, a solute molecule may be associated with several solvent molecules, but the attaching forces in these groupings are much smaller than those between atoms in a molecule, or between melecules in a crystal. For example, Al(OH)3 normally exists in a water solution with three water (H2O) molecules attached to it, too.
In a colloidal dispersion, the dispersed "particle" is MANY MANY molecules bonded tightly together - almost as tightly as in the original material from which they came. It's just that the original mass of material has been broken up into VERY tiny particles. Milk, fog, and whipped cream are good examples. Milk has both fat and protein particles suspended in water, along with some truly disolved solutes like trace metals. And note that milk, as it comes from the cow, will separate if undisturbed. The fat portion will float to the top as "cream", and eventually some of the proteins will sink to the bottom. But most milk we see has been homogenized. That is, the suspended particles have been modified to make them repel each other so strongly they refuse to collect in one area, and the colloidal dispersion is said to be stabilized. Fog is water droplets (each containing thousands (billions? more?) of water molecules, with the droplets suspended in air as the "medium". Whipped cream is air droplets as the colloidal material, themselves suspended in a colloid medium called cream (see milk above).
Regarding light dispersion, light will be scattered off the interface surface between any to media with different refractive indices. And since EVERYTHING has its own refractive index, there's always a difference IF the particles in contact are large compared to the light wavelength. For a true solution the solute individual molecule is MUCH smaller that the wavelength of visible light, so the light simply does not interact with it as a particle with a different refractive index. (There are other types of interaction - for example, abosrption of visible light by dye molecules - but that's another story.) But for a MANY-molecule colloidal particle, that particle is large compared to the light wavelength and the light "sees" an object very different from its surroundings; hence it is scattered at the surface. Thus, one old rule of thumb is that light will pass easily through a true solution, but will be scattered by a colloidal dispersion.