A construct is a theoretical abstraction of something that can't be measured directly, e.g. happiness. Construct validity addresses whether measurements intended to relate to that construct, e.g. smiling, can actually be used to make valid inferences about the construct.