Both social network analysis and ML interpretation are associational in nature, not causal.
E.g.: Do more LinkedIn posts cause more funding? Or does more funding cause more LinkedIn posts? Or does a confounder, like better sales or other metrics, cause both more LinkedIn posts and more funding?
Your post (and especially the Sifted article) implies the first causal structure. But I highly suspect it's a combination of the latter two.
Also, conditioning on an outcome (success) additionally creates collider bias.
Interesting finds, if a bit fallacious IMHO :)
Both social network analysis and ML interpretation are associational in nature, not causal.
E.g.: Do more LinkedIn posts cause more funding? Or does more funding cause more LinkedIn posts? Or does a confounder, like better sales or other metrics, cause both more LinkedIn posts and more funding?
Your post (and especially the Sifted article) implies the first causal structure. But I highly suspect it's a combination of the latter two.
Also, conditioning on an outcome (success) additionally creates collider bias.
I.e.: P(tech degree | startup success) =/= P(startup success | tech degree) =/= P(startup success | do(tech degree))
I see people making these spurious correlations so often, it kinda irks me haha.