The following excerpts from Toronto newspapers indicate the reactions expressed following the failure of a Tandem `non-stop' system at the Toronto Stock Exchange August 16th. Note that the Tandem system did indeed not stop, but the three disc drive failures prevented access to critical market information, without which trading could not continue. "TSE `crash' drives trades to Montreal, Brokers curse computer system" (Mark Hallman, Financial Post, Thursday August 17, 1989, p. 1.) A computer crash all but shut down trading on the Toronto Stock Exchange for almost three hours yesterday, forcing tens of millions of dollars' worth of trades to Montreal. ... [the crash] -- a multiple failure within a disc-drive subsystem -- forced a halt at 9:41 AM. ... `Two pieces of hardware break down and Bay Street breaks down.' said a sour Charles Mitchell, trader. ... `Who's accountable for that?' ... A TSE spokeswoman said the failure of both primary and backup systems had never occurred since computers were installed 26 years ago. ... `Today, the market floors are so automated that when you have this kind of computer failure, there is nothing that can be done', exchange President Pearce Bunting said. `The greater your dependence on computers, the greater the risks.' The exchange's computer assisted trading system (CATS) was not affected and smaller stocks continued to trade. But only about 25% of the stocks listed on the TSE are traded using CATS. ... `Everybody's very much annoyed,' McLean McCarthy's Mitchell said. `It's costing us a lot of money. I think the people upstairs in the exchange should be held accountable for it.' [Note: Exchange floor traders failed to haul out the traditional chalkboards to continue trading manually "in the interests of a fair market".] "Fail-safe Tandem system fails" Geoffrey Rowan and Lawrence Surtees Globe and Mail, Aug. 17 1989, p. B-1 The stock market crash of 1989 -- yesterday's three-hour trading halt at the Toronto Stock Exhange -- was caused by multiple failures in a computer [Tandem VLX] designed to work even when some of its pieces are broken. ... Peter Richards, president of Tandem Computers Canada of Markham, ont., said the disk drives suffered three separate failures that were a `once in a lifetime event'. ... Mr. Richards said yesterday's failure was not a computer crash. `The whole system, minus the data on the one drive, was up and running and available to the exchange during the entire morning. That also meant that recovery was done with the computers on-line, making it a quicker process than with standard computers." "TSE seeking to prevent further computer chaos", Geoffrey Rowan, Globe & Mail, August 21, 1989, p. B-1, B-4. ... `Fault-tolerance guarantees that you will not have a problem from any single failure', said John Kane, vice-president for marketing strategy for Tandem. `I've never heard of three failures at one one time. I don't know what the gods were doing to us.' [P. H. Roosen-Runge, Dept. of Computer Science, York University, Toronto, Canada]
I've been watching the stream of messages deriding as hazardous, impossible, stupid, etc., etc., the idea of automatic steering of cars on the highway with some bemusement - and concern. The purveyors of the lastest bit of technical wizardry are always ready to call it "absolutely safe", "foolproof", and so on. In response, we seem to be developing a "nay-saying" attitude: Any use of computer controls is inherently unreliable, risky, even downright dange- rous. This "see no good" response is no better than the "see no evil" atti- tude of the purveyors! Let's try to think about the problem of controlling automobiles RATIONALLY for a moment. This is a problem which, WITH SUITABLE DESIGN, is MUCH better solved by machines than by human beings! Driving on a highway consists of long periods of boredom punctuated by random, unpredictable moments in which quick responses are required. People are very bad at dealing with this kind of situation. They get bored, their attention wanders; few are adequately trained to respond quickly, without hesitation or thought, to the near-instan- taneous (on the human time scale) dangers which can arise. Computers, on the other hand, cannot get bored, cannot start thinking about their date that evening, and can respond very quickly, without hesitation or panic, if something goes wrong. It is quite true that a system which mixed human with machine drivers would be very hazardous: Neither is good at dealing with the responses or the abili- ties other. In any case, I'll bet that MOST of the hazards of highway driving are the result of problems with the human drivers, not with the road, the mechanical parts, or other external factors. In some 20 years of driving, I've had my share of scary highway encounters. Without a single exception, they were all clearly the result of poor human decision-making: Unexpected lane changes without checking whether the lane was clear; driving way too fast for the icy conditions of the road and going into a spin at 60 mph; wandering to the edge of the road, touching the curb, then panicking, losing control over the car, and wandering over several lanes of traffic before regaining control. And then of course there are the drunks and the druggies, who account for a large percentage of accidents. It is also quite true that the highway system as it is designed today was intended mainly for human use. Lane markers that are so clear to the human visual system are notoriously difficult for computers to see, for example. But if the goal is a computerized highway, there is no reason in the world to limit yourself to systems that make sense to humans. A lane marker is hard to track, but a buried wire can be an absolute triviality. Do you worry about the computers controlling elevators? Elevators have been controlled by digital computers - albeit relay-based - for 30 years or more. They, too, are mechanical devices subject to all sorts of failure. A computerized driving system, designed for such a purpose, operating in a constrained environment, strikes me as quite practical. It would have to be designed carefully, with "fail-safe" modes - which should be doable since, while tasks like finding the best route to a point 100 miles away require global information and coordination, failing safe requires local information (where are the cars around me, what do their computers claim they intend to do in the next 100 ms., what do my sensors show they actually ARE doing (an essential, and fairly simple backup) and so on.) Any system can be implemented badly. The structure of some systems makes bad implementations almost inevitable. Computer control of cars on today's road system, in combination with today's drivers, would be such a system - and it seems to be the system most people are commenting on. But it is NOT the system we must necessarily build! -- Jerry
If one were to compare the potential level of hazard, based, say, on fatalities/100,000 miles, which might arise from failure of on-board computers as opposed to the current levels of fatality caused by intoxicated and/or drugged-out drivers, heart attacks, strokes, vehicular suicides, driver error, police chases, etc., I wonder which methodology (hardware vs. wetware) would actually be safer, once all the emotional rubbish about "men vs. machines" was cleared away? Just a thought...
Bring back the trains, buses and streetcars. These are well-proven, reliable, non-polluting ways of moving large numbers of people from place to place. My intuition is that the problem of controlling thousands of automobiles from a central point would rival Star Wars in complexity.
Don't worry too much about these. If we can't get people to ride public transportation (like buses and subways) then they won't sit still for the equivalent in their own cars. And anyway, how do you get off the "remoteway" to stop at the cleaners? Emily H. Lonsford, MITRE - Houston W123 (713) 333-0922
>Call me a technophobe if you like, but *NOBODY* can guarantee 100% ... Well, it must be nice to live in the UK. Here in the US we have a very reliable manual system. It can be so relied upon to kill 40,000 people a year that we simply take it for granted. A large percentage of these deaths involve malicious manipulation of the system, i.e., the ingestion of alcohol and other drugs. You may be able to tolerate these deaths in preference to the risk of computer systems that are less than perfect. Quite candidly, I am sick to death of a system that tolerates and defends this carnage, while crying crocodile tears about computer-related risks. It is high time that we got a system for comparing risks that is prepared to put these lives on the same scale that we use to condemn research into alternatives. The New York Times concluded in a front page article that the American people are far more likely to tolerate the huge death toll from their discretionary and recreational use of highways, tobacco and alcohol than risks which are comparatively minuscule from automated systems. I may be stuck with that, but it is insane and elitist. It is high time somebody said so. William Hugh Murray, Fellow, Information System Security, Ernst & Young 2000 National City Center Cleveland, Ohio 44114 21 Locust Avenue, Suite 2D, New Canaan, Connecticut 06840
This has been brewing for a long time. I enjoy the RISKS forum and tout RISKS as the best digest available on e-mail/netnews. But sometimes I wonder. Submitters seem to take great glee in presenting YAHS (Yet another horrible story), but I seldom see any constructive comments -- or much discussion at all -- mostly it is simply story time. And most of the stories are about why we should not trust technology. But the unaided human is even less reliable than the aided one! Surely we could use the stories as design examples and attempt to use them to inform our design, so that we could invent technologies that were real improvements. Instead of using the examples for a cheap laugh, how about using them for instruction? Let me provide some examples, using RISKS DIGEST 9.14 as my example, both because it is handy and also because it had an excess of scare stories and cheap conclusions. - - - 1. Misdelivered mail because the barcode had the wrong zipcode, and the machines read it and ignored the written address. Moral: I would say that you should not use code that was only machine readable. Either do like the banks do -- invent a machine-readable text that humans can also read -- or do as supermarkets do - print the "English" version of the bar code beneath the bar code. This would make it a lot easier to catch errors of this sort. (The best solution is to make postal machines that can read the handwriting on the envelopes, but that is another decade away, at least). (Because barcode is already standardized as the postoffice system, why not require English text to be printed beneath the barcode? No change in equipment is required at the Post Office end.) - - - 2. Automatic vehicle navigation systems. RISK readers are really scared of this one, to judge by the number of times this is raised in RISKS. But what are the alternatives? Today we have roughly 50,000 fatalities a year in the auto. If driving were all done automatically, there would be occasional massive screwups. But would the death toll hit 50,000? I bet it would improve things. (Yes, roughly 1/2 of those deaths have drinking implicated, but I don't see how this changes my argument.) These systems in aviation are reasonably efficient. Moreover, I see no reason why we couldn't make these things fail reasonable softly, with lots of warning. And yes, there would still be foul-ups, but the baseline is not zero accidents -- it is 50,000 deaths). (Admission of guilt: I am consulting for a company that is designing one of these automated systems.) - - - 3. One submitter's statement that "I am amazed that Boeing has taken all the blame... " for pilot's shutting down the wrong engine. The submitter than cites other examples of pilots shutting down wrong engines, implying that this is clearly pilot error (since it happens often and in various aircraft). So Boeing -- or any particular manufacturer is clearly not at fault. I disagree. This error has long been noted, first well documented in the 1940's. But it is the design of the cockpit controls that leads to the error. So I would still blame Boeing (or cockpit designers in general). I have long thought that many control panels are designed as if to increase the chance of error (my experience is mostly in nuclear power and aviation), and that there are redesigns -- perhaps using better technology -- that would minimize these errors. Shutting off the wrong engine is so common (relatively speaking), that more design effort should have gone into finding better systems. Instead, folks just blame the pilot. That is the wrong attitude. Look, people make errors. Fact of life. Therefore, good design will anticipate those errors and make them either impossible or less likely, or easier to detect and correct. (Admission: I am off to Boeing next week to discuss cockpit design. This note will not win me friends at Boeing.) - - - 4. The report of airliner crew fatigue. Cockpit crew fall asleep and so cabin crew are now supposed to wake them up, or there are systems that blast them with loud sounds. My solution is quite different: let them sleep! It is well known that the circadian rhythm has two minima -- about 2PM and 2AM. And flying long hours across time zones is hard on the system. Suppose you let crew take naps. As long as at least one member stayed awake, and if the nap were announced and known about, there is likely to be no problem. Modern aircraft can be flown with one crew member: more are needed only in emergencies and in periods of high workloads (e.g., takeoff and landing or in crowded airspace at low altitudes). In fact, some airlines are experimenting with allowing short naps. The point is, human biology should be lived with, not faught against. Naps are natural, and overall efficiency will be improved if you allow very short naps. And I suspect we can develop schemes whereby this improves safety, not hinders it. (Citation: Somewhere buried in my office is a reference to a paper on this topics I just heard at a meeting of the (British) Royal Society on Human Factors in High Risk Technologies.) - - - FINAL STATEMENT: My argument is that human physiology and psychology are reasonably fixed: we have to design systems to take account of our properties. But too many of our technologies are designed with only the technology is mind, and the human is forced to cope. Then when the problems ocur, we blame either the techcnology isteslf or the person. But neither is to blame -- it is the design and use of the technology that is wrong. We could design technologies that inproved our lives and considered our strenghts and limitations. That is what I am after! (I was told just yesterday by a visiting NASA researcher that modern Air Traffic Control Centers are clustering operators based on considerations of cable lengths and cabling neeeds, even though this is non-optimal in terms of operator performance -- people who must communicate with one another are thereby put far apart. It is this attitude I am fighting -- it isn't the technology that is wrong, it is the lack of consideration of the human side of the operation that is faulty.) People are especially good at some things, bad at others. Creativity, adaptability, flexibility, and extreme reliability are our strengths. Continual alertness and precision in action or memory are our weaknesses. Why not use technology in ways that take advantage of the former and allow for the latter? And in reviewing the cases presented in RISKS, why not use them as guides to better designs and better technology, not as excuses to lash out against the use of technology in general. Don Norman INTERNET: email@example.com Department of Cognitive Science D-015 BITNET: dnorman@ucsd University of California, San Diego AppleLink: D.NORMAN La Jolla, California 92093 USA [By the way, even in RISKS Volume 1 there was a complaint that RISKS dealt only with problems and not with solutions. I have on several occasions noted that RISKS is limited by what is submitted, and that very few people have ever submitted success stories. On the other hand, I frequently quote Henry Petroski's statement that we learn NOTHING from our successes and at least have a chance to learn from our failures. But the real -- and not even subliminal -- purpose of RISKS is to make us all aware of what the risks are so that we can intelligently try to avoid them! No disagreement there! PGN]
Computer Ethics: Cautionary Tales and Ethical Dilemmas in Computing by Tom Forester and Perry Morrison Published by MIT Press and Basil Blackwell To Appear (Hopefully) January 1990 CONTENTS Preface and Acknowledgments 1. Introduction: Our Computerized Society Some Problems Created for Society by Computers - Ethical Dilemmas for Computer Professionals and Users 2. Computer Crime The Rise of the High-Tech Heist - Is Reported Crime the Tip of an Iceberg? - Targets of the Computer Criminal - Who Are the Computer Criminals? - Improving Computer Security - Suggestions for Further Discussion 3. Software Theft The Growth of Software Piracy - Revenge of the Nerds? Intellectual Property Rights and the Law - Software Piracy v. Industry Progress - Busting the Pirates - Suggestions for Further Discussion 4. Hacking and Viruses What is Hacking? - Why Do Hackers `Hack'? - Hackers: Criminals or Modern-Day Robin Hoods? - Some "Great" Hacks - Worms,Trojan Horses and Time-Bombs - The Virus Invasion - Ethical Issues - Suggestions for Further Discussion 5. Unreliable Computers Most Information Systems are Failures - Some Great Software Disasters - Warranties and Disclaimers - Why are Complex Systems So Unreliable? - What are Computer Scientists Doing About It? - Suggestions for Further Discussion 6. The Invasion of Privacy Database Disasters - Privacy Legislation - Big Brother is Watching You - The Surveillance Society - Just When You Thought No One was Listening - Computers and Elections - Suggestions for Further Discussion 7. AI and Expert Systems What is AI? - What is Intelligence? - Expert Systems - Legal Problems - Newer Developments - Ethical Issues: is AI a Proper Goal? - Conclusion: the Limits of Hype - Suggestions for Further Discussion 8. Computerizing the Workplace Computers and Employment - Computers and the Quality of Worklife: `De-skilling' - Productivity and People: Stress, Monitoring, Depersonalization, Fatigue and Boredom - Health and Safety Issues: VDTs and the RSI Debate - Suggestions for Further Discussion APPENDIX A Autonomous Systems: the Case of "Star Wars" Preface and Acknowledgements The aim of this book is two-fold: (1) to describe some of the problems created for society by computers and (2) to show how these problems present ethical dilemmas for computer professionals and computer users. The problems created by computers arise, in turn, from two main sources: from hardware and software malfunctions and from misuse by human beings. We argue that computer systems by their very nature are insecure, unreliable and unpredictable - and that society has yet to come to terms with the consequences. We also seek to show how society has become newly vulnerable to human misuse of computers in the form of computer crime, software theft, hacking, the creation of viruses, invasions of privacy, and so on. Computer Ethics has evolved from our previous writings and in particular our experiences teaching two courses on the human and social context of computing to computer science students at Griffith University. One lesson we quickly learned was that computer science students cannot be assumed to possess a social conscience or indeed have much awareness of social trends and global issues. Accordingly, these courses have been reshaped in order to relate more closely to students' career goals, by focussing on the ethical dilemmas they will face in their everyday lives as computer professionals. Many college and university computer science courses are now including - or would like to include - an ethics component, but this noble objective has been hampered by a lack of suitable teaching materials. Computer Ethics has therefore been designed with teaching purposes in mind in an effort to help rectify the shortage of texts. That is why we have included numerous up-to-date references, as well as scenarios, role-playing exercises and `hypotheticals' in the `Suggestions for Further Discussion' at the end of each chapter. The creative teacher should be able to build on these. Readers will notice that we have not adopted an explicit theoretical framework and have avoided philosophical discussion of ethical theory. The reason is that this book is but a first step, with the simple aim of sensitizing undergraduate computer science students to ethical issues. Neither will readers find a detailed account of the legislative position around the world on the various topics discussed. This is because in each country the legal situation is often complex, confused and changing fast - and again this is not the purpose of the book. Finally, a note on sources. First, we have to acknowledge an enormous debt to Peter G. Neumann, whose "Risks to the Public in Computer Systems" sections in Software Engineering Notes, the journal of the Association of Computing Machinery's Special Interest Group on Software (ACM-SIGSOFT) have provided inspiration, amusement and a vast amount of valuable information. Long may he continue. Second, we have to caution that many of these and other sources are newspaper and media reports, which, like computers, are not 100 per cent reliable. Tom Forester, Perry Morrison School of Computing & Information Technology Maths, Stats and Computing Griffith University, University of New England Queensland, Australia Armidale, NSW, Australia
Please report problems with the web pages to the maintainer