When we have a curve contained in a immersed surface we can define two notions related to the usual curvature of a curve and torsion of a curve. If we name to the unit normal to the surface and to the product , we can define:
Another approach (with other notation) (@needham2021visual page 116), if the curve is and the tangent vector at the point of study is we can consider the normal to the surface , and the planes (the tangent plane to ) and the normal plane to the surface spanned by and .
Suppose the osculating plane to at is inclined an angle ( is the angle between the binormal of and the normal to ). Then and are the curvatures of the projections of onto and respectively. If the curvature of is it is satisfied that
As might be expected, the geodesic curvature can be obtained intrinsically. The inhabitants of the surface do not know that they live in a surface so they can try to measure the usual curvature by their own means, for example, by the measuring the change in the angle, as is outlined here.
They weren't using tangent lines, indeed, but "tangent geodesics". The result is: what for them is the curvature, for us is the geodesic curvature.
See @needham2021visual page 119.