Subject: Hazards of Airliner Computerization (Brint Cooper, RISKS-9.16) >Such reasonableness checks as humans would be capable of performing, >would be far from "make work" and would reduce significantly some of the >risks associated with increasingly automated flying. Such reasonableness tests in software are not easily constructed. It is much more difficult to do than you seem to think. Subject: Computers in Medicine (Brint Cooper, RISKS-9.16) > 1. Is my perception correct? Are there proportionally more > life and property threatening computer-related faults in banking, > transportation, and national defense than in medical applcations? No. No, there are not. There are proportionally the same given the relative complexity of the systems and the amount of use. > 2. If there's even a modicum of truth in #1, then why? There is no truth in it. > 3. Or are the physicians merely burying their mistakes again? Be careful to have your facts before attacking a group of people. In fact, medical computer problems are more carefully reported because the FDA requires this while there are not similar requirements in other applications. There are large numbers of reported computer errors and recalls in the FDA database. I have noticed, however, that many medical equipment manufacturers take quality control more seriously than some other industries because of the potential liabilities and costs involved in an FDA-ordered recall of a medical device.
Don Norman (email@example.com) writes: > 4. The report of airliner crew fatigue... My profound apologies for omitting part of the BBC report; I typed it in from memory. Don's statements reminded me that the BBC report included considerable discussion on the subject of allowing aircrew napping. The basic argument was generally that which Don raised, i.e., that minor aircrew napping would actually make things safer. It did, however, sound like the "powers that be" were rather reluctant to allow the public to know that the folks up front may be sawing logs. As an aside, a few weeks back NPR had an extensive report on napping. The contention was that studies conclusively show that napping dramatically improves concentration, productivity, and the ability to deal with problems. Yet society strongly disapproves of napping, and anybody caught napping on the job will be very lucky if they don't get fired. Napping is viewed as "wasted time"; time which could be spent doing something "constructive." Never mind that major portions of most workers' afternoons are often spent in an advanced state of bleary-eyed semi-coherence. It will be a long time before this attitude changes, despite the fact that some of history's most effective personalities, such as Winston Churchill, took frequent naps. Michael Trout, BRS Information Technologies, 1200 Rt. 7, Latham, N.Y. 12110 (518) 783-1161
Don Norman wrote: >... It is well known that the circadian rhythm has two minima -- ... 1. It's extremely difficult to apply simple principles of the analysis of circadian rhythm when the crew in question might be transiting multiple time- zones, on a continuous basis, on irregular schedules. 2. Part of the problem is cockpits that lull their pilots to sleep. In _The Journal of Navigation_, an oft-repeated statistic through 1983 and 1984 was that Britannia Airways found that 767 cockpit crews were far more sedate (lower heart rates) than those flying the 737. This was ascribed to better-designed seating, noise reduction, and low workload. In cruise, modern crews are given very little to do. To address this problem, very positive steps are being taken--such as having the stewardess check in, airline policy requiring manual calculation of navigation problems, mandatory (unnecessary) radio call-ins every fifteen minutes, etc. I see no reason to criticise such measures: they attempt to make the best of what is becoming increasingly clear is a bad situation. 3. Operational practice on many airlines (even if it's not *policy*) IS to let a pilot take a snooze if he just can't keep his eyes open. The other pilot's notified, and the guy takes a nap. Great. But that reduces the redundancy in a modern cockpit by half. So what happens if the remaining guy (often on the same sleep-schedule) falls himself nodding off? (if you've never flown long-distance, it's quite startling, after being *awake* 25 hours or so, to suddenly wake up--without ever discovering you had fallen asleep). Modern autopilots are incredibly reliable, and it's easy to argue that there's nothing wrong with even *both* pilots napping for a while. The problem is that, even in cruise, even with modern flight assist mechanisms, problems can manifest themselves faster than the crew can "get back into the loop." One of the rec- ommendations of the NTSB report on the China Airlines flipover over Los Angeles a few years ago was to bring pilots closer into the control loop. My personal gripes with flight automation are: * Lack of experience. Automation's a new concept. We have little experience dealing with the human factor in automation design. * Economics-oriented vs. safety-oriented design. It's always possible to fund a study that supports the manufacturer's viewpoints. Even if backhanded tech- niques aren't used, genuine divisions exist within the industry on the place and role of automation. * Lack of human-factors influence in system design (letting the engineers have at it). * Too much human-factors influence in system design (psychologists pushing pet theories in safety-critical situations). (the above really boils down to lack of *operations* influence in system des- ign, but even that's not always a help--consider a system designed to pilot specifications, which leaves the pilots with precious little to do). * Relatively low lack of computer literacy among pilots (the more microproc- essors the better), and consequent over-reliability on automation (Cf. my article in RISKS 9.13). * Unfounded fault-probability claims by manufacturer (both in program ver- ification and hardware reliability). * Lack of standardization. How one airplane does something is not necessarily how another one does it (look at airspeed displays on the 747-400, Fokker 100, A320, and A310). Not even COLORS are standardized. In my opinion, not nearly enough muckraking is being done on the risks of automation in aviation (and other disciplines). When computers are marketed on the mere basis that they're high-tech, and purchased, respected, and put in safety-critical situations solely on that basis, it's time to worry. I think RISKS serves a purpose by bringing such problems to our attention. And I, for one, do not "get a laugh" out of reading about how yet another implementation has been botched up.
Another opinion on "drive-by-wire" systems: I, like many other submitters, am against them, but for a reason not yet brought up. Designers of automated traffic-control systems have an unfortunate tendency to design for cars, and forget about other road users, like cyclists and pedestrians. A case in point is the "smart" traffic light, which is actuated by automotive traffic. There are a lot of these menaces where I live. They work fine for cars, sure. But try to make a left turn when you're riding a bike. You get in the left-turning lane, right on top of the sensor, in hopes that it will notice you. If you are riding a steel bike, you have about a 70% chance that the light will turn for you. The sensors, I believe, work by magnetic inductance, so if you're riding an aluminum bike (as I do) you have a less than 50% chance that the light will turn for you. Of course, if the light doesn't turn, you are forced to run it. This is more dangerous than if there were no light there, because the drivers going straight through think they have the right of way, and aren't expecting turning traffic. Another problem for cyclists is that "smart" lights often go green for only an instant, so that if the road is crowned, it's difficult to get through the intersection while the light is still green (or, in some egregious cases, while it's still yellow). These smart lights often endanger pedestrians, too. A block and a half from my office is a "smart" T intersection. For a long time, the light was adjusted so that left-turning traffic and pedestrians would be in the same place at the same time. (I think the designers forgot that this was a T intersection, so that all the traffic would be turning left.) Again, it was particularly dangerous because the drivers thought that the pedestrians were jaywalking, when in fact the pedestrians had a green light. Finally, after at least three years, the problem was "fixed" by barring pedestrians from crossing there at all. Technology lovers might argue that these are design flaws that could be fixed. This is true, but they haven't been. (I, and other members of my bicycle club, routinely call the cities and couties when we find lights that don't turn for us, but we rarely get satisfaction. And why should it be up to us to make sure the lights work, anyway?) I don't believe that "drive-by-wire" systems would be any better. By the way, where I live, the number of cyclists is not insignificant. In Santa Clara County, twice as many people commute by bicycle as commute by mass transit. I would be happy to hear from Donald Norman, or other people working or consulting for companies that are designing "drive-by-wire" systems, on how such systems will allow for cyclists. -- Anne Paulson, Intellicorp, 1975 El Camino Real, Mountain View, CA
My long diatribe in RISKS has generated the proper result: numerous people have written me and several, as this piece from Anne Paulson indicates, have elaborated in the spirit of my remarks. Although my piece in RISKS argued that technology was not all bad, my main goal in life is to convince the technologists to consider the human side of things. There is a tendancy to let the technology dominate, forgetting the inconveniences this causes for people. How do I convince anyone? Well, I write, preach, and evangalize. In my consulting, my entire emphasis is on ways of understanding the needs of the users, and then finding ways to bend the technology to fit the people (instead of the other way around). Consider Paulson's point, that designers of traffic lights (intersections and traffic flow) do not properly consider pedestrians and bicyclists. Actually, I think she underestimated the problem: I suspect these designers consider pedistrians and bicyclists a nuisance that are best gotten rid of. The best approach I ever saw to these concerns was in a letter to the editor in a local nespaper. The author asked why we had "pedestrian crossings" at streets. Shouldn't pedestrians come first? We should have "car crossings." That is, the pedestrian should automaticlly be considered to have the right of way, and cars would have to have special places where they could cross the pedestrian stream. The same argument goes for bicycles. Paulson concludes by saying: "I would be happy to hear from Donald Norman, or other people working or consulting for companies that are designing "drive-by-wire" systems, on how such systems will allow for cyclists." That isn't the right approach. The correct approach is to change the entire mindset of the city planners and people who purchase these devices to put in the proper emphasis. The engineers and designers are often very good, with proper motives, but they can't overcome the mindset of cities that compute traffic flow, and purchase from the lowest bidder -- even if the equipment thus purchased is inferior and the system considerations neglected. [In addition, I think that automobiles, bicycles, and pedestrians have such different characteristics that they simply should not be in the same traffic streams. We need special bicycle ways (separated from roadways by more than a painted line) and special walkways. (Some European cities -- espcially Scandavian cities -- seem to take this approach.) And we should separate diffrent kinds of vehicles as well. And as we get more and more elderly drivers who tend to drive slowly, we will need either efficient point-to-point mass transit or special driving lanes for these elderly "super cautious" drivers (reaction and decision times slow with age, and attention is limited, with less ability to divide attention among several tasks. Those of us in the attention business sometimes say that the elderly have "less attentional resources.").] The problem faced by RISK readers, designers, and users, is that society puts cost and efficiency first, and cost and efficiency are measured by local, monetary variables. Real cost and eficiency would take into account accident rate, long-term pollution, long term recycling, long-term learning, and employee comfort and job satisfaction. In the end, I think proper attention to these factors increases morale, decreases sickness, increases efficiency, lowers turnover (which thereby lowers training costs), and lowers the cost of cleanup and technological fixes by society. But society is driven by short-term views. It will be hard to change this mindset. But I am optimistic. Don Norman, Department of Cognitive Science D-015, University of California, San Diego, La Jolla, California 92093 USA
firstname.lastname@example.org quotes a recent article by Carl Levin. Briefly, "Airlines are starting to fly a new generation of highly automated jets, raising concerns among safety researchers that pilots will rely too much on the technology and will lose or never learn the sharp skills and reflexes needed in emergencies." Most of the article quotes the fears of "experts" and cites recent examples where a pilot's skill overcame emergency conditions, saving many lives. However, the article does not objectively cite balancing situations in which automated systems handled sets of conditions that were too complex and changing too fast for humans ever to respond satisfactorily. The article includes an irrelevant reference to the old saw that calculators prevent pupils from learning the "basic principles of mathematics," neglecting the fact that one does not do "mathematics" on calculators. One does "arithmetic" on calculators! Only near the end of the article are cockpit simulators discussed. "A pilot in a simulator can practice flying after losing various computer and control systems...Still, while simulator training could help for some kinds of emergencies, others... are considered so remote that pilots do not train for them on simulators." One wonders, "Why not?" Can the great minds of the aircraft industry not postulate virtually any kind of emergency? Can not the simulators be designed/programmed to handle any kind of emergency? This sounds very much like a failure of will, not of ability. Much of the problem expressed is not new. For years, the US Air Force has been flying high performance tactical jets using significant amounts of automation. If the automated system fails, the game is over. The demands made upon the pilot/airplane system are too demanding ever to be met by humans. These pilots are among the best in the world, and their training heavily depends upon cockpit simulators. The entire article is a good example of the warped treatment afforded a complex, scientific topic by the popular press. That example, how the public is told about these issues, is probably the best contribution made by such an article to the dialog in the Risks Digest. Perhaps, instead of writing to one another so much, the well-informed among us might write to such as the NY Times, giving more accurate pictures of the issues. _Brint
After reading much discussion on the subject, I'd like to call to attention the cancelled "Skybus" project that would have replaced the old streetcar lines in Pittsburgh. [A feature story on the subject was aired on KDKA-TV Eyewitness News last night; some of the facts presented here came out of the story.] 20 years ago, Westinghouse designed and built a prototype "Skybus" system, which involved driverless, electric rubber-tire vehicles on an exclusive track. This system was billed as being the cheapest and most efficient public transportation system available. However, many politicians and citizens were opposed to the idea of robot vehicles operating on tracks, "not being able to tell whether an object on the road is a newspaper, a concrete block, or a human body". Of course, this doesn't take into account the fact that the roads were elevated and, for the most part, isolated from the rest of the world. As a result of the public hysteria, the system was canned, and the new LRV system was built, at five times the cost of Skybus. There is no doubt in my mind that such a system can and will work, with the proper safeguards taken into account. By isolating the track, there is little need to risk collision with humans, although there would be a need to scan for foreign objects that could serve as obsticles. As for the spacing and switching of vehicles, this is already an automated function for many railroads. What recent railroad accidents can be directly blaimed on computer failure? The point of this story is that the problem can go both ways. We can easily get burned by relying too much on automated systems that can be prone to failure. By the same token, we can get burned economically by refusing a technological advancement that might present a few risks, but at the same time would solve many others (as I feel the Skybus project would have done). In fact, since then, Westinghouse has teamed up with a German firm (the name evades me at the moment), and has implemented Skybus-like systems in many locations, including Morgantown, WV. In fact, the new mid-field terminal at Greater Pitt. Intl. Airport will be implementing such a system when the terminal opens in '92. George (HAL) Feil, Carnegie -Mellon University bitnet: gf08%andrew@CMCCVB
`Drive-by-wire' - YES i agree that the `human controlled' system causes fatalities at an unacceptable level - YES i agree that the druggies/alcoholics should be kept off the roads, YES i agree that there is a place for civilised public transport (tramcars, buses) and YES i agree that theres a place for automated navigation systems. BUT i don't think they will solve the worlds transport problems. Most journeys (in Europe at least) are of less than 10 miles - taking the kids to school, collecting groceries, popping out for a pizza, and it is the case in the UK that most accidents happen within 5 miles of your home. These are the very situations where the `blocks' of automated vehicles travelling close up, just will not work - they are in most cases urban areas with too many hazards (stop lights, school buses, intersections). The instances when the bunched, automatic vehicles WOULD be of use are on long-distance journeys. Paradoxically, on a basis of fatalities per thousand miles, the motorways, autobahns and autoroutes of Europe are the safest roads. For the long journeys, there are trains and planes. Having only been to the States once, and only driven on minor roads, i can't really get a feel for Stateside freeway conditions. But how can a `civilised' country have a 55MPH limit? Come over to Europe, drive round the London orbital motorway (where the speed limit is 70MPH, but the police turn a blind eye to anybody doing less than 100) and sharpen up your driving skills! The human brain, if it is working properly, is still the best real-time adaptive guidance system we have. Whats more, it comes fitted as a standard, no-cost feature in most people. Its catastrophic failures are thankfully rare, and if it does suffer one, then you aren't going to be worrying any more in any case! Pete L.
I would like to echo some of the sentiment expressed by Tim Shimeall in RISKS 9.13. I am a non-degreed "software engineer". I started programming as a kid in 1968 and although the term was not used at that time, I was a "hacker". After graduating high school in the early seventies I began attending a community college, the best that I could afford. Unfortunately the draft was on and when I got that compelling notice I enlisted. I honed my craft while in the service, bypassing the service's programmers school because of very high test scores. Within six months, I was appointed the system administrator at a scientific research and development facility. When not handling system matters I was tasked with helping others with their debugging problems. Interestingly enough, there were very few "uneducated" enlisted folk at this site and most of the college trained junior officers were my best clients. They could not program their way out of the proverbial paper bag. When I decided to leave the service, it was decided that my job would be converted to a GS civilian slot. I applied for the job and was turned down because of my lack of a college degree. The base personnel office told me that I was unqualified for a job that I had been doing so well that I was recommended by my supervisor and selected for commendations, as well as NCO of the quarter. I eventually went to work for a large computer vendor who sold me back to the Government to fill my old job. They got $100,000/year for my services and the facility management people who I worked for as an NCO were thrilled with my encore performance. In the past twelve years in this business I have held the following titles: Computer Programming Specialist, Systems Analyst, Senior Systems Analyst, Associate Software Engineer, Software Engineer, Senior Software Engineer, Principal Software Engineer, Computer Security Engineer, Systems Engineer, Senior Computer Scientist. I have always felt proud of the work that I have accomplished. I am a published author and have attended invitational conferences. I have always been employed based on my reputation, not on my formal education. I have managed people with far more formal education (and less talent) than I. I have tried to obtain a degree over the last twelve years but it has not been an easy road. My work has caused me to travel a great deal and to relocate often. At one point I was asked to work abroad for a couple of years. Each time I have moved I have attempted to matriculate to a new university only to lose a substantial number of credit hours to a version of the "not invented here" syndrome. This has made obtaining my "credentials" not only near impossible by very costly. I have taken three different courses in "operating system design" from three different schools and gotten an "A" each time. In one course I corrected the professor's misstatements (discreetly of course) about the operation of the virtual memory demand paging algorithms of a system that I helped to maintain. Don't get me wrong, I feel that formal education is a worthwhile experience. But, for those who must pursue this science without the benefit of a college education, self education is a viable alternative. I think the software engineering community places too much stock in formal education and not enough in the proven ability to do the job. As Mr. Shimeall pointed out in RISKS-9.13, apprenticeship has long been recognized in this country as an extremely reliable measurement of one's ability to perform. An office mate of mine who holds an advanced degree in computer engineering once told me that obtaining a degree just proved that you could "play the game". I apologize for the personal nature of this transaction but Mr. Jones' comments brought to a head the frustration that I have felt for a number of years. It seems that people in this field are more interested in where you went to school than in what your opinions on various design issues are. I remember an incident a couple of years ago when I submitted a paper to a conference. It was reviewed and I received a letter of acceptance in the mail along with a request for a biography to be published with the paper. Shortly after sending my biography I received a curt reply explaining that my paper was being dropped from the schedule and would not be published. The explanation was that this was a "professsional" conference and surely I would understand. My protests went unanswered. I was recently rejected voting membership to the IEEE because they questioned my professional status because I was non-degreed. I guess my point is here is that the degree doesn't make the engineer... -ear
Please report problems with the web pages to the maintainer