This recent research by Lawrence Zitnick on a handwriting averaging approach based on sampling is not only interesting, but holds enough potential to influence the way we do type design. Basically, the algorithm takes user stroke samples in real time (with the help of a pen tablet) to increase stroke consistency in a non-destructive manner. Take a look at the presentation video (link at the bottom); there's no need to put it into words.
I wonder: If this algorithm could take stroke-width information into account, the application to the digitization process of calligraphic ‘handwritten’ fonts is obvious.
Putting sampling aside, the paper suggests more: an algorithm could take an existant font, analyze frequent characteristics among all different glyphs (within the same font) and average independently over each this characteristics. This makes sense specially because the algorithm doesn't destroy variation, as it always averages: tendencies are reinforced. Like the idea behind the average font, but built upon a much more clever procedure.
Even more helpful would be the partitioning of glyphs* into primary components (stems, bowls, serifs, terminals, etc.) to allow global tweaking of weight, contrast and proportions without making use of interpolation. Of course, this is a naïve statement, but it isn't necessarily unfeasible.
Link to the presentation video:
Link to the paper:
* independently of the style of the typeface.