I read a Re-Twitt from our good blog friend Abel Pharmboy of Terra Sigillata which pointed to a most amazing bit of online drug information. The originator was @mjrobbins of The Lay Scientist blog*. The answer to "What are the most addictive drugs" on some Q/A site called blurtit. And dude, do they have answers! For example, the "addiction likelihood" of cocaine is 78%, of heroin is 87.5% and of crystal meth is 89.5%. Wow, what precision! And they cite some science-y sounding dudes including some that I recognize as being experts. Great stuff, all kinds of people want to know the objective addictiveness of different drugs right? From parents to policy makers, users to scientists.
...but what's all this? The "addiction likelihood" of cannabis is 42%, of LSD is 32% and of psilocybin and mescaline are 16-18%? Hmm.....
Luckily the answer on blurtit contains citations to their sources. And tracking back to one of the data sources, we find a clue. . At least some of the backing data are derived from a study described as follows:
"To rank today's commonly used drugs by their addictiveness, we asked experts to consider two questions: How easy is it to get hooked on these substances and how hard is it to stop using them? Although a person's vulnerability to drug also depends on individual traits -- physiology, psychology, and social and economic pressures -- these rankings reflect only the addictive potential inherent in the drug. The numbers below are relative rankings, based on the experts' scores for each substance:
Hmm. And nicotine is at 100 so in the parlance of the blurtit answer, nicotine is "100% addictive". Which is patently false. This blurtit answer fails to communicate that the numbers given are subjective rankings and relative rankings at that. I think a very large swath of people would read that blurtit answer and infer that these numbers are intended to reflect some sort of objective, absolute "addictive" quality of the drugs.
As my readers know, getting at questions of "addictiveness" and conditional probability of dependence (as I like to frame the question) is a tricky business and it is essential to tie the seemingly objective rankings / percentages / whathaveyou to the underlying analysis or data. So that the reader can infer "If you look at these numbers, the answers are..." instead of "This is something which reflects fundamental truth..."
Additional data sources cited in the blurtit "answer" seem to refer back to a similar ranking procedure described here. You will notice these data use a 1-6 scale ranking, not a percentage. Of course deploying these particular data might give the reader a clue that what was being discussed was relative addictive potential, not some sort of objective truth. If you tell someone a drug has addictive likeliness of "5", they are quite naturally going to ask "5? Relative to what?". If you tell them it is "85.5%", some large fraction of the audience is going to assume something or other about the underlying population- any users, consistent users, 'those losers', etc. But they are not going to automatically assume "...on a subjective opinion scale of a limited number of experts wherein the most addictive substance is set to 100%".
It is a horrible answer to the question.
*Updated, forgot to include this link and h/t when originally published.