Reverse of the law of large numbers

I found this fascinating piece of news on FT today. Summarising the idea:

In general, the law of large numbers states that large sample size leads to convergence towards expected value. Reverse LLN, on the other hand, describes the reverse effect of LLN. In a market where there are many participants, herding behaviour reverses the LLN effect, moving the probability distribution away from normal towards the extreme tail end. Therefore, market views become unbalanced, tipping on either end of the extreme. This has implication on market prices, enabling prices to move away from equilibrium and hence, driving bubbles formation in the markets.

http://www.ft.com/cms/s/0/2005e96a-33f1-11dd-869b-0000779fd2ac,s01=1.html?nclick_check=1