« eBay v. MercExchange: A Brief Analysis | Main | Discouraging Patent Holdouts »

May 16, 2006


Feed You can follow this conversation by subscribing to the comment feed for this post.


"The police can’t be everywhere, we rarely know the people driving near us on the freeways, and this combination of rare surveillance and practical driver anonymity contributes substantially to aggressive driving. Largely as a result, vehicular collisions are the leading killer of Americans aged 15 to 29."

What does this mean? That aggressive driving is the largest (or one of the largest) causes of collisions? Cause you'd probably want some kind of evidence for that, if so.


Yes, that is what it means.

I provide extensive citations in support of that factual statement (and the other statements in my post) in the paper.


Interesting paper, though I only had time to skim through it. My one thought is that implementation may be difficult because of incentives for other drivers to contribute their opinions.

In Ebay, the exchange is mutual--you usually leave feedback for people who leave you feedback. This is far from perfect because of the prospect of mutually assured destruction creates the incentive to only leave feedback when it is positive and you expect positive feedback for yourself. As you point out in the paper, such strategically malicious feedback can ameliorated somewhat by programs designed to ferret out deliberately bad feedback, but given the small reward of leaving any feedback in these systems, even the most remote chance of loss in the form of bad feedback may deter many potential participants from leaving bad feedback. The familar volunteer problem may deter people from leaving any feedback at all--but if you pay them that distorts the population leaving feedback.

The opposite problem may also occur due to the lack of motivation--other drivers may be motivated to leave feedback when the conduct is egregiously bad. By and large this will not be a problem, except that this means a single reported incident will cause severe consequences (much as on Ebay a single negative rating causes severe consequences for most participants). This amplifies the effect of any errors.

William Rothwell

In the section on whether to provide feedback, you make a few suggestions on how to decide which feedback gets passed along. One recommendation you make is to evaluate "driving expertise" via the point system and pass along the advice of the "driving experts." Taking a cue from some of the online review networks, I would add that it might be worthwile to evaluate "feedback expertise" and then pass along the advice of the "feedback experts."

For example, on Amazon.com there is both a review of the item (CD/book/etc.) and then a review of the review. The first person types up their thoughts on the item and posts the review along with a 1-5 star rating. People browsing these reviews are given the option to answer the question "Was this review helpful to you?" by clicking yes or no. Amazon then sorts the reviews based on how helpful people thought they were such that the most helpful review is posted first.

A similar approach might work here. When a person hears their feedback, they can be given an opportunity to report whether it was useful. Through this mechanism, we can identify those individuals who, for whatever reason, are particularly good at communicating feedback. The HMDFE system can be designed to give preference to these "feedback experts." This could be used in conjunction with the driving experts rankings so that after we muzzle the non-top quartile, we can then rank the remaining drivers on their feedback expertise to select out the drivers that are experts in both driving and feedback-giving. Alternatively, the system could be set up to provide a mixture of expert driver and expert feedback-giver opinions.

Two final points:

1) There might be some worry that, unlike the Amazon system where the reviewer tends to provide a positive service to the feedback evaluater, under the HMDFE system the recipient of feedback might resent the feedback-giver and have motivation to inacurrately appraise the quality of the feedback. I would be interested to know if Amazon has any system in place that prevents the author of a book from finding positive reviews and repeatedly ranking them as helpful (and ranking negative reviews as unhelpful) so that prospective purchasers see the high-ratings first and have to dig to find any negative opinions. The situations are not identical, but it might provide some guidance for dealing with the "bad-faith rater" problem.

2) It might be worthwhile to consider the value of providing a monetary incentive to the highest rated feedback-givers just as we are, through lower insurance costs, providing a monetary incentive to the highest rated drivers. This provides not only some incentive for all drivers on the road to pick up the phone and call (rather than ignoring the bad driver), it also provides an incentive to give useful feedback. Further, because it only rewards the best callers providing the most useful feedback, it seems unlikely to incentivize malicious feedback. Drivers calling in for malicious reasons will, presumably, not be able to leave the kind of helpful feedback that would result in a positive rating.


How about a mandatory "How's my writing?" section on academic papers, along the lines of this comments section? John Donohue recently presented a paper at the law school in which he suggested that the academic work on the effect of the death penalty has been misleading. He would presumably like to alert readers of those studies to their deficiencies. Some of the professors who attended his talk, meanwhile, would most likely want to alert readers to deficiencies in Donohue's paper.

Judges would probably find such a system unconstitutional. This raises the obvious question: Should judicial opinions have mandatory "How's my Constitutional analysis?" comment sections?


Have you ever noticed how everyone, even the worst drivers among us, think that they're great drivers? And have you noticed that most people personalize everything - i.e. a guy might be driving down the road talking on his cell phone, swerving out of his lane, doing 10 miles under the speed limit, but if you honk at him when he almost hits you, he has the gall to make an obscene gesture at YOU?

Most drivers have no business rating the driving skill of other drivers.

I havent read your paper, but it sounds like the ebay comparison quickly breaks down - on ebay, 2 people mutually decide to enter into a transaction, each with the understanding that the other party will then be entitled to submit feedback regarding the transaction. In your example, however, anyone and everyone can leave feedback for anyone else. Do you mean to say that by driving my car onto the road i expressly agree that every person on earth has the right to submit an opinion about my driving skills?

You'll note that ebay expressly disallows this sort of thing - only parties who have engaged in a transaction with one another may leave feedback for the other person.

Imagine that your tenure was tied to a popular vote - anyone who wanted to vote could vote, regardless of whether they have any legal training and regardless of whether of they'd actually read your paper. I havent read your paper, and I have my uninformed opinion on the premise. Should my opinion have any bearing on your salary or tenure status? Should my mother's?


The main problem is that, in general, people in America do not see breaking traffic laws as a "real crime". Everyone breaks the speed limit, many don't use turn signals, some don't turn their lights on in the rain, etc.

I really think Americans just need to change their attitudes towards driving. We need to make it clear that driving is a privilege, not a right. We could start taking away driver's licenses more often (for traffic violations), it only makes sense. People who have proven that they cannot drive safely don't have any right to drive IMHO.

People who get their licenses taken away will tell sob stories about needing to go to work and drive the kids around etc., but if you have all of that responsibility, there's no way you should be driving irresponsibly. When a person negligently endangers many other people's lives, they don't deserve our sympathy.

Raise the stakes, and our streets will be safer in no time. The threat of driver's license suspension will coerce most people into driving by the rules (which are designed with safety in mind, and are thus justified).


Thanks for these comments. I'll respond in order.

TJ raises a terrific point. As the cost of feedback drops, the leaving of feedback will begin to look more and more like eBay's reciprocal system, which can only be healthy. For that reason, much of the design challenge in a universal How's My Driving? program is to make the act of leaving feedback as inexpensive as possible.

William also raises terrific points, and it seems to me that rating of feedback helpfulness could be incorporated into the system that I propose, subject to the dangers that he flags. I'll try to say more about this in the final draft of the paper. Thanks.

James -- We already have something like a "How's My Writing?" system. For blog posts, they consist of comments like the ones you left. For judicial opinions, they consist of academic commentary and judicial decisions from other jurisdictions. Draft papers get revised based on sensible comment submissions, and judicial doctrines sometimes get overruled if academic or judicial commentary is uniformly negative. I'll blog more later about "How's My Driving?" for Everything, but you've hit the nail on the head by noting the possibility of further applications.

Shack - Members of the public, by and large, lack the expertise to say anything meaningful about whether a faculty member should be tenured in law or physics. Students, who watch us teach and perhaps read some of our scholarship, will have some expertise and their sentiments are given weight in tenure decisions. Senior faculty members have the expertise to evaluate the quality of scholarship, and their "feedback" is given a great deal of weight. The difference is that most drivers on the road are perfectly capable of identifying safe or unsafe driving, even though they are incapable of differentiating between, say, good and bad linguistics scholarship.

Freethinker - Your comment suggests that police enforcement of traffic regulation is optimal. I spend a lot of time in the paper explaining why that's probably not the case. There are some dangerous forms of driving that police are good at catching (e.g., speeding) and other dangerous forms of driving that police are bad at catching (e.g., tailgating or excessive braking). I am arguing that the current system overdeters the former and underdeters the latter. So your sentiments are reasonable, but my paper is designed to change your mind.


"The difference is that most drivers on the road are perfectly capable of identifying safe or unsafe driving, even though they are incapable of differentiating between, say, good and bad linguistics scholarship."

On this, we disagree. Simply because a person drives doesn't make him an expert on identifying unsafe driving behavior. that's like saying that any person who has ever taken a course, at any level, is qualified to critique a harvard professor's teaching style - they're not.

As I've already mentioned, most people are oblivous to their own actions and how their actions affect others. If people are capable of identifying unsafe behavior, then why do so many drivers continually engage in unsafe driving practices? Are they all irrational actors? For your theory to be valid they must be. If so, who cares what they think about other peoples' driving skills?


"How's My Driving?" programs for commercial vehicles successfully identify the worst drivers based on phoned-in feedback. Those drivers are then re-trained or fired, resulting in substantial reductions in accidents. If your intuition were correct, then "How's My Driving?" placards would be ineffective. As it happens, they appear to be quite effective and a multi-million dollar industry that relies on that effectiveness has emerged, in the U.S. and abroad. Unless you think that people are capable of identifying bad truck drivers but not bad automobile drivers, your intuition is inconsistent with the available data.


""How's My Driving?" programs for commercial vehicles successfully identify the worst drivers based on phoned-in feedback."

that's quite a statement.

a program like this identifies drivers and then labels some of them "the worst." simply applying a label doesn't make the label fact. the fact is, you have no idea whether these drivers are in fact "the worst." all we know is that the system identifies them as such.

you point to a handful of sources that claim a roughly 25% reduction in crashes after the implementation of an HMD program. i'm not blind to data, but i have to say, i am pretty skeptical - one the the sources lists a sample of 445 trucks. i don't know the accident rate of truckers, but i can't think that it's high enough that a 22% reduction in crash rate is statistically significant.

also, the implementation of any HMD program is almost certainly accompanied by other safety-inspired changes at the company. this alone may account for the cited difference between the apparent effectiveness of HMD programs and black-boxes.

finally, as you admit in your paper, none of these studies has been peer reviewed. some of the cited stats even come out of trade magazines; knowing something about that industry, i can tell you that trade mags are desperate for content and will print almost anything.

in sum, i take issue with the veracity and impartiality of these statisitcs, and very much doubt that HMD programs are effective at all.

to tell you the truth, ive lost interest in this topic, so i'm not going to comment here anymore. but i will check back later, should you choose to respond. thanks for the discussion.


As you know, the study of 445 trucks indeed found a 22% reduction in the crash rate, but it also found a 52% reduction in crash costs in the year following implementation of a "How's My Driving?" program. That's one of several event studies mentioned in the paper, so I'm not sure why you want to pick on one 445-n study to indict the entire literature.

In any event, I am troubled by the lack of peer review for these studies, and I say so in the paper. I reported the best evidence I can find. But to go from there to doubting "that HMD programs are effective at all" is weird. In addition to the data, we know that within a few years, the number of trucks sporting HMD placards has grown substantially, and large trucking companies have shelled out hefty sums to the firms that offer HMD-monitoring services. Maybe these trucking companies have been hoodwinked or value the P.R. benefits of HMD placards greatly, but either of these scenarios would be surprising.

Ben Redmond

A national "How's' my Driving" system would not do any good. There are already license plates on every vehicle that legally traverses the roads and highways. If you wanted to you could have dialed 911, told them it was a non-emergency and filed a complaint about the reckless or rude driving, giving them a description of the vehicle and it's LISCENCE PLATE NUMBER.

The License plate number is already in place for this and other reasons, so why would we waste the time and resources trying to set up another system for identifying vehicles and drivers? Would the "How's my Driving" sticker be easier to see then the License Plate Number? Would it be easier some how for the Police to find them? Or was it just soo frustrating an experience that you had to dream up a way of trying to find them and teach them a thing or two?

The system in place for identifying the drivers who offend is fine as it is, if those who see the offense occur would just speak up to the authorities.

Oh, by the way. I am in the transportation business in Seattle and I know for a fact that the two biggest motivations for a company putting "How's my Driving" comment systems stickers on their trucks is 1) It is GREAT p.r. to make it look like you care about the safety and feelings of the public at large, who in noticing the sticker also notice your company logo and 2) It is a threat (usually false), to the employee driver, if they are the only one who will be driving that vehicle.

It would be a far more laudable and effective thing to try and reach those who are driving bad and convince them of the benefits they could personally experience if they slowed down. "Drive Write"


The license plate argument, though valid, seems to be off point. There are many virtues to the "How's My Driving?" sticker that the license plate cannot offer. First of all, drivers themselves will likely drive more safely when they have the sticker on their cars than they do without it. It seems to me that having a sticker that gives other drivers a direct line for reporting bad driving will result in more calls/reports than expecting other drivers to call the police and give them a license plate number. The license plate does not effectively curb reckless driving because people know that other drivers are not likely to call the police to make a report, and even if they do, there are probably not going to be any serious consequences. Second, I do not think Lior's point (perhaps I am wrong on this one) was to have drivers call the number and then have the police run out and stop somebody. If we stick with Ben's license-plates-are-fine system, this is what we will have. When a driver calls the police to report reckless driving (which probably does not happen as much as we would like to think, since the license plate is not very inviting to this option), if the police want to do anything about it, they will have to rush out and try to catch the reckless driver in the act. With the sticker system, we can keep a database of feedback on a person's driving habits (we can even keep feedback on the feedback - see William's post above) and then use this to give out warnings and even possibly as a factor in calculating automobile insurance rates. Third, Ben asks why "we would waste the time and resources trying to set up [the sticker system]?" It seems to me that the system would actually save time and resources. It would have the same function of giving drivers an outlet to report reckless driving without needing to have our police officers rush out to the scene every time. Their time will be saved and their manpower can be used elsewhere; all this while we can still keep records of reckless driving that can be used effectively in many ways.

State Rep

I read the paper and I'm impressed. Shack must think insurance companies give discounts for unproven programs. What I'm not sure about is how politically viable this proposal would be.

Paul Farrell

I was a study sponsor within the insurance industry (Fireman's Fund Insurance Co.). Our initial study included 30,000 power units from 200 fleets spread across the country in a variety of industries ranging from Pest Control (pickup trucks and smaller vans) to delivery operations (creameries, 800-flowers, etc.) The program worked for us (22% reduction in collisions over the time period prior to using a hotline service). Why did it work? There were few if any consequential changes to the safety programs at the firms in the study group during the trial (three and a half years of data). Those who coached and trained drivers saw the most significant improvements - judged by their report closeout rate and by their actual claim history. If no training or closeout of report, then there was virtually no change in loss activity (some but it could have been luck). The key, in my mind, was the re-training effort made by the sincere users of the program. Since that study was done, about a dozen more insurance studies have been documented - not just the Hanover study with 445 trucks. Additionally, large private companies with thousands of trucks have studied the reduction - still in the 25% range or better. I am in the business now as a supplier so you may challenge my bias, but I took the job as vendor because I believe it works. I don't think HMD for everyone will work - no re-training component ---its not just the notice, its not embarrasment, its follow up by someone who has authority to get driver to think about behavior, consequences and motivate a change. We try to save lives, not get people fired, but its not our call as vendor to dictate the outcomes.

Paul Farrell

One point about anonymous reporting - the National Transportation Safety Board (who investigate tragic crashes) have previously issued recommendations that anonymous HMD reports should be investigated as they may signal a trend in behavior and should not be discarded outright because they are anonymous (this was in a report on the BurntCabins, PA crash of an interstate motor coach where the driver drifted off the highway and crashed the bus).

Dorob Regev

This is a great article and Lior did a very important work. Nevertheless the point made about the luck of research is very true.
Not less important is the question of what to research! In other words if reports are taken and passed on to management - but if management does very little with them if at all, the reports alone will not do much for safer driving.
Israel is the first country we know of who introduced a regulation requiring all commercial vehicles to enroll in the HMD system. Because there are large fleets here of high tech workers, There is a huge and very significant number of small vehicles driving with the famous bumper stickers. A system we developed allow for employees to receive the reports directly to their inbox (as well as replying by email directly in to the database).

We are looking forward to work with anyone and share our unique software and experience.

Doron Regev Co-Ceo
Tnuanet, Israel
[email protected]


Why did it work? There were few if any consequential changes to the safety programs at the firms in the study group during the trial (three and a half years of data).

Paul Farrell

Why did it work with few if any changes to safety programs? They simply used the programs that they had in place, but focused more time and attention on the drivers who were the subject of the reports from HMD. The existing programs worked better -- rather than train everyone in a bulk environment, spend more one-on-one time with those few drivers who actually demonstrated a need for re-training.

The comments to this entry are closed.