Assuming you are referring to the standard style on opensreetmap.org: the rendering of individual elements colour, patterns, labels etc. is defined in a stats static style file, that contains rules for each zoom level. See https://github.com/gravitystorm/openstreetmap-carto for the source code. You cannot influence the cartographic decisions by object tagging, nor would that be a sensible thing to implement. If should be noted that with so called "vector tiles" which are rendered in your browser instead of as bitmaps on a central server, there is a bit more flexibility. However for practical reasons there is still a zoom level based object selection for the contents of the vector tiles.

Assuming you are referring to the standard style on opensreetmap.org: the rendering of individual elements colour, patterns, labels etc. is defined in a stats style file, that contains rules for each zoom level. See https://github.com/gravitystorm/openstreetmap-carto for the source code. You cannot influence the cartographic decisions by object tagging, nor would that be a sensible thing to implement. If should be noted that with so called "vector tiles" which are rendered in your browser instead of as bitmaps on a central server, there is a bit more flexibility. However for practical reasons there is still a zoom level based object selection for the contents of the vector tiles.