I've seen people touting "collaborative diffusion", in reference to this paper. However, I am having trouble understanding the benefits. How is this different than an influence map? If it is a useful technique, what is the best way to implement it? The paper seems to be mostly patting themselves on the back for not falling into the OO Object = Real World object trap, but doesn't talk much about actual implementation.

I've seen people touting "collaborative diffusion", in reference to this paper. However, I am having trouble understanding the benefits. How is this different than an influence map? If it is a useful technique, what is the best way to implement it? The paper seems to be mostly patting themselves on the back for not falling into the OO Object = Real World object trap, but doesn't talk much about actual implementation.

Yep. Influence maps and yet the word "influence" only appears once in the entire paper. I suppose it looks really impressive because he's using squiggly letters for his diffusion function. I've seen a few others like this specifically on influence maps. It's obviously someone who stumbled upon what he believes to be a novel solution because he is cloistered away from the standard game AI literature.

Oh well...

What's with the bizarre 3D plots? He probably thinks they look spectacular, but a simple heat map would have done a much better job of conveying the information.