Category Archives: justice

Safer Streets Through Social Disruption

The Star has been bludgeoning the Tories’ plans for crime and punishment in Canada all week. In summary, being “tough” on crime is expensive, ineffective, and creates more of the social disruption in which many crimes already have their roots.

I’m not going to deny that there are some people who are just broken. They are psychologically unable to exist in society, no matter what situation they find themselves in – good or bad. But the vast majority of criminals aren’t pathological axe murderers. There are a lot of people who made bad choices in times of stress, to pay off crushing debts, for example, or feed their families. In these cases, a job might be all it takes to get these people back on the straight and narrow.

The whole matter of incarcerating people who shouldn’t really be in jail for years got me thinking about two concepts used in scientific research that are applicable here: sensitivity and specificity.

Sensitivity is the ability of a test to correctly identify positive cases. For example, if 100 people commit homocide in a year and the justice system puts 95 of them behind bars, the justice system has a sensitivity (with respect to homocide) of 95%. This also means that 5% of murderers go free.

Specificity is the ability of a test to correctly identify a negative case. For example, if 100 innocent people are accused of murder in a year and 95 go free, the justice system has a specificity (with respect to homocide) of 95%. This also means that 5% of innocent people accused of murder are convicted.

I borrowed this chart from Wikipedia:

Actual State of Guilt
Guilty Not Guilty
Verdict Guilty True Positive = 95 False Positive = 5

(type I error)


= TP / (TP + FP)
= 95 / (95 + 5)
= 95 / 100 ≡ 95%

Not Guilty False Negative =5

(type II error)

True Negative = 95 = TN / (TN + FN)
= 5 / (5 + 95)
= 5 / 100 ≡ 5%

= TP / (TP + FN)
= 95 / (95 + 5)
= 95 / 100 ≡ 95%

= TN / (FP + TN)
= 5 / (5 + 95)
= 5 / 100 ≡ 5%

Even in science it is nearly impossible to establish a test with 100% specificity and 100% negativity. The better you get at sending all guilty people to jail – approaching a sensitivity of 100% – the more likely you are to imprison an innocent person – specificity falls. The same is true vice versa. If you want to ensure an innocent person never goes to jail, you have to accept that some guilty people will slip throught the cracks.

It’s been a while since I needed you use sensitivity and specificity, so I’ve almost certainly made a mistake in the numbers (maybe even the definitions) but the principle remains. Would you rather see all guilty people punished or all innocent people unpunished? We can’t have it both ways.