Wednesday, November 14, 2012

One Thing Leads to Another

Some time in the 1980s when I was teaching a big introductory computer programming course, one of my TFs brought to my attention two programs that seemed to be copies. They were submissions by two students in his section. I asked the two students if they could explain the similarities and they claimed they could not. They had not communicated or discussed their papers. In fact, they had never met each other and neither even recognized the other's name.

Students confronted with their sins often initially deny wrongdoing (and for that reason I no longer surprise students with such accusations -- I email them and ask them to come see me, so they can get past the denial stage while sleeping on it). But I thought for these two students to deny that they even knew each other was pretty bold. It gave me pause. As I am wont to do when puzzled, I brought it up in TF meeting, showed my assistants the two papers and asked them how they thought we should handle the situation. 

A couple of other TFs recognized the submissions as matching ones they had received. It turned out there was a chain -- everyone had been careful not to cheat with someone in their own section, as they knew their TF would spot the similarities immediately. They had not calculated that as the chain grew it might reconnect. The two original students may well have been telling the truth about not copying from each other.

Today such assignments are cross-compared automatically and not only are cheaters revealed, but the full structure of any web of influence is disclosed in an instant. 

I thought of this story while reading the New York Times news analysis this morning, Online Privacy is Also In Play in Petraeus Scandal. As the report states, 
The F.B.I. investigation that toppled the director of the C.I.A. andhas now entangled the top American commander in Afghanistan underscores a danger that civil libertarians have long warned about: that in policing the Web for crime, espionage and sabotage, government investigators will unavoidably invade the private lives of Americans.
Most of us are now used to (even if, apparently, not fully conscious of) some of the unusual properties of bits: that they stick around indefinitely, that they can be moved and stored at almost no cost, that they can be searched easily. And I am glad that the public is now suddenly aware of the low level of legal protection for your email, if you use a cloud service like Gmail.

But this case makes a somewhat different point. Social connections tend to be numerous and many of them weak. Gumshoe investigations used to require a lot of effort to track down who knew whom who knew whom. Some human judgment was required to discriminate between potentially strong or suspicious connections; a chain of three or four weak connections would get you so far from the origin and into a circumference of such huge size that it just wasn't worth following all the leads.

No more. The full social web of any of us is a scary thing. Even people without infidelity issues have second-cousins-by-marriage involved in sketchy activities. Do we all need to ask ourselves, for example, whether we should have been more careful about refusing small favors from distant relatives? Must every molehill be viewed as a potential mountain?

So I am glad to see that people are asking how it was that half a dozen mildly annoying emails from one woman to another led to the resignation of the CIA director, and what that says about our expectations for government officials and others. Let's get real: we can't keep losing good people this way, foolish men though they be.

But I also think Harvard is going to have to ask itself how much it really wants to know before it launches another "cheating" investigation like the one involving Gov 1310. According to the Crimson, this started when the professor reported 10-20 students to the College, which subsequently decided to compare all the hundreds of take-home exam papers to each other. It seems this was done by hand over the summer, not electronically, but the fact that the papers were easily retrievable was likely a digital phenomenon. I am told by relatives of students in the course that until fairly recently, the College was still contacting students with new examples of allegedly suspiciously similar papers submitted by other students.

One can see the scope of this case as evidence of a deep rot in the moral fabric of Harvard students, or even college students in general. I tend to see it more as the result of faculty negligence. Perhaps we will be able to make a clearer judgment about all that when the College finishes its work and reports the body counts.

In either case, it illustrates, as the Petraeus case also illustrates, the risks in the digital era of asking too many questions while in a self-righteous frame of mind. Why not have automatic cheat-detection software review all papers submitted to all courses, or all email sent by top government officials? Because the world has ambiguous and evolving standards (and while Gov 1310 seems to have been egregious, I doubt that precision about collaboration is likely to become universal any time soon). In a world of imprecise standards that require human judgment to apply fairly, the full social cost of surveillance, the collateral damage in shattered dreams of relatives and friends, and the intimidating effect on the community of making plausible but unwise accusations, will be greater than the benefit to be gained from the attempt to keep society squeaky-clean. We need to remember that just because digital technologies have given us tools to do important investigative work with unprecedented efficiency and thoroughness, that does not mean we should use those tools to the full extent of their capabilities. Such tools also allow us to create destruction at a scale that used to be impracticable if not impossible.

No comments:

Post a Comment