The general rule of thumb is never to introduce a degree of complexity you don't need.
Use the expected behaviour of the structure to determine the element type. A two-way slab is usually best modelled as 2D elements (plate elements) and an orthotropic bridge deck is often best modelled using 1D elements (beam elements) because it consists of elements which only have one span direction each.
This makes the model easy to create, and what is perhaps even more important, it makes the model easier to check. The closer the model is to a manageable hand calculation method, the more realistic it is to do spot calculations that verify the results of the model.
The use of 3D elements may also be problematic because the software will often assume the material is able to transfer the same stress in each of the three directions. Rolled steel is an example of a material which isn't isotropic and may have very low tension capacity in the thickness direction due to various defects (lamellar tearing). The easy fix is to only use 2D elements that will automatically assume zero stresses in the thickness direction.
For reinforced concrete, making a design based on 3D elements will be a bit a nightmare unless you make an extremely detailed model that include the reinforcement and take cracking of concrete in tension correctly into account, because otherwise you will have tension stresses in concrete that exceed its tension capacity and you will have to convert the stresses in the 3D elements to equivalent section forces corresponding the results of 2D elements in order to do a reinforcement design. To correctly account for concrete cracking in three dimensions is a complicated problem and to attempt this in a model of the global behaviour of a structure is a challenge that is best avoided if at all possible.
In most cases, I will only recommend using 3D elements for two situations:
Local models of complex details in steel structures and similar. This includes cases where you to need to know the exact peak stresses to account for fatigue effects.
Soil in geotechnical problems with significant 3D effect.