Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
When Computers Rule
“I wanted to kill myself. If I hadn’t been pregnant, I definitely would have killed myself.” Thus, does Seema Misra begin the story of her three-year odyssey as a sub-postmistress working for Britain’s Royal Mail.
She was approved to open a postal station in her small Surrey shop in 2005, and was set up with supplies and a Horizon computer terminal to use for Post Office business. (A word to the wise: the Post Office in the UK operates differently than it does in the States, and there are over 11,000 of these “sub-post offices,” usually operating out of small general stores, chemist shops, or newsagents, dotted all over the country, often in out-of-the-way villages (think Agatha Christie mysteries) which would otherwise be underserved. The people who run them are contractors, not employees of the Royal Mail. And the post office itself is used for purposes unheard of in the US–you can go there to pay your utility bill, perform banking transactions, and get your welfare payments, as well as buy stamps and send off letters and packages.)
It wasn’t long before Seema began to notice discrepancies in her end-of-day reconciliations, between what she thought should have been the total Post Office business done, and what the computer report spat out. In every case, the computer report indicated that there should have been more money applied to the account than indicated when Seema took the receipts and transaction history and totaled them up manually. Some days, the discrepancies were in the £100 range. Sometimes, thousands.
She reported this as a problem both to Horizon (Fujitsu) and to her supervisors at the Post Office. They advised her that she was liable for the discrepancies because she’d signed a contract promising to make good any losses. So she began feeding profits from her store into the Royal Mail system to make up the difference and balance the Post Office account. She kept reporting the problem. They kept insisting the trouble was with her.
She was suspended from her job with the Royal Mail in January of 2008, when an audit found a discrepancy of £74,000 in her accounts (about $100K in today’s money).
And after a court summons and a guilty verdict, Seema Misra was sent to jail where she spent months in the company of the dregs of society. While she was on the inside, her husband was repeatedly beaten up by neighbors who called his wife a thief. The family lost their home, and her record as a felon made it difficult for them to rent, and subsequently for her to get another job.
This year, for the first time, and as the result of an October 2020 statement by the Royal Mail that it will “not contest” the efforts of people such as Seema to get their convictions overturned (“Bally decent of the old chaps,” I can just about hear Bertie Wooster saying), there is hope for Seema and her family.
There are over 1,000 people like Seema Misra in the United Kingdom, victims of what some have called “the worst miscarriage of justice in the legal system in modern British history.” It’s an unbelievable horror of a tale, and I won’t belabor it (there are a number of links at the bottom of this post, if you’d like to read more), but in a nutshell, the facts are these:
- In 1995, the Royal Mail instituted a pilot program at several offices involving a computerized “smartcard” system to automate the payout of welfare benefits to prospective recipients and reduce fraud (LOL).
- The rollout failed (imagine my surprise) and the project was scrapped after an expenditure of about three-quarters of a billion pounds, but from its ashes arose something called “Horizon,” a point-of-sale (POS–LOL again) system for Royal Mail transactions.
- Problems with balancing and reconciliation were quickly noted and publicized.
- The IT vendor (Fujitsu) and the Post Office repeatedly insisted that each case was unique, that “no-one else” was reporting similar problems (this was false and they must have known this at the time), and did nothing to help its contracted employees.
- The Post Office failed to find a problem with the software, commissioned an audit, canceled the audit a day before it was due to be published, and concluded that there was no systemic problem with the Horizon system.
- Between 2009 and 2013, the Post Office began to admit that, yes, there were bugs in the system, but that the system was “working as designed” and was fit for purpose. It strenuously denied reports by sub-postmasters and postmistresses that it appeared as if transactions were being altered “after the fact” by Fujitsu technical support, and insisted that such things simply weren’t possible.
- Investigative audits began to track errors in the software, including the fact that it wasn’t tracking certain transactions, was recording some transactions in duplicate, and was disadvantaged in some cases by old or inadequate equipment. The Post Office dismissed most of these claims, insisted that the problem was inadequate training, and that, instead, the bulk of the problem lay with the sub-postmasters and postmistresses who were, in a word, thieves.
- At some point, the audits and investigation concluded that it was, despite the Royal Mail’s insistence to the contrary, entirely possible that employees at Horizon/Fujitsu could have intervened and changed the data unbeknownst to the postmasters/mistresses, and that, in fact, they likely had.
- An organization, Justice for Subpostmasters Alliance, got things somewhat organized, and secured backing to reduce the fear (which many of the individuals had) that they couldn’t possibly contest the charges because if they did they’d, by law, be required to pay the Post Office court costs. This may have been the moment when the tide began to turn.
Meanwhile, The Royal Mail launched an aggressive campaign against people reporting the discrepancies and difficulty balancing, and over the course of several years, hundreds were sent to jail. Hundreds were disgraced. Many were placed on suicide watch. At least one committed the act. Hundreds of singles and couples who’d taken on the Royal Mail commission as a fillip for their retirement income were embarrassed, humiliated, shamed, and disgraced. In many cases, they were jailed and permanently branded “thieves.”
In December of 2019, in a blistering 400-page ruling, a judge ruled that “bugs, errors and defects in the Horizon system was the cause of the discrepancies which had ruined hundreds of people.” (550 of them were part of the class-action lawsuit which led to this ruling.) He also opened the door to the idea that the software defects should allow the defendants/convicted felons the right to petition to have their guilty verdicts overturned.
And that is what has led to the Royal Mail’s generous decision that it will “not contest” the efforts of people like Seema Misra to get their lives back after more than a decade in Hell. Jolly big of them. (Each “convict” has to petition individually, and have the case heard and the verdict rendered.)
Meanwhile, there’s a £58 million class-action settlement which, by the time all the fees are paid, means that participants in the suit will receive a pittance for their victimization, bullying, and terrorization by the all-powerful State.
It’s an absolutely sickening story. And now, for the rest of it. Warning–Strong opinions follow:
Almost nothing fills the heart of the person in IT-world with dread so much as the thought of being on the receiving end of a barrage of criticism from hundreds, maybe even thousands, of the great unwashed in Realville, complaining or explaining why the marvelous and perfect system he envisioned, coded, tested, and filled with the bells and whistles of his dreams, isn’t satisfactory and may even–quelle horreur, c’est impossible!–have a few defects. Not only is there the obvious ego problem, there’s often a language problem as well, as the computer-illiterate (not in any way intended as a slur) struggle to communicate with someone speaking from the rarefied heights of Technology Privilege.
It’s not dissimilar to, and I think the chasm is about as wide, the language difference between the sexes. I am sometimes reminded (as I often seem to be) of an old Punch cartoon from the late 1950s or early 1960s. (Probably the latter, since I must have been old enough to appreciate it and remember it.) A well-dressed lady–think Maggie Thatcher, or Margo Leadbetter (for fans of the old BBC series “Good Neighbors“)–is trying to explain to the garage mechanic, in as much detail as she can, exactly what’s wrong under the ‘bonnet’ of her vehicle. He’s standing there in his filthy overalls, scratching his…umm…belly and rolling his eyes, and probably imagining what fun he’s going to have telling this story to his mates at the pub, and she’s saying, I’m sure in a very Received Pronunciation sort of way: “It sounds like hairpins rattling around inside a plastic cup.”
The first response of our hero in IT-world (I was one, so absolute moral authority, and yes I’m exaggerating a bit for effect. But those of you who’ve lived the dream, tell me if I’m not right over the target) to such a presentation, and to an often inelegant and inaccurate attempt to describe the problem, is to try to get rid of these people as quickly as possible so he can get back to WoW or whatever was occupying his time prior to the nuisance call, and so she can get back to her knitting. There are a few tried and tested responses to help this along:
- “Is that so? I’ve never heard of that before.”
- “I can’t replicate your problem.”
- “You must be doing it wrong.”
- “No one else is reporting your problem.”
- “Who told you to do that?”
And finally, the big guns:
- “You must be mistaken. That’s just not possible. The computer isn’t wrong.”
Case closed. One of, or a collection of, these responses will probably get rid of more than half of the first-time callers who slink back to their desks in shame and promise themselves never to try something like that again.
If the person in Realville is persistent, isn’t intimidated into silence and a sense that she must have screwed up somehow, and if she really believes there’s something wrong (and not with her), where does she go next? To her supervisor, of course! Except that he’s probably not much of a computer person either, and there’s nothing he wants to hear less than that the system he and his company recommended and spent upwards of £1 billion to purchase, and years to implement, isn’t doing the job and may actually be distorting and corrupting the books. A moment’s thought, and it will occur to our doughty corporate warrior that calling the complainant a liar or a thief, and ordering her to make up the difference from her own pocketbook will probably shut her up, and if it doesn’t, at least it will move the problem out of his jurisdiction and into someone else’s.
And so they did.
Full disclosure: I knew nothing of this story until last week, when my sister and I were discussing the daily sausage-factory of new reports over possible election malfeasance in the United States. I remarked that “nothing would surprise me anymore.” And I also (full disclosure again) said that as a person of IT privilege (as both I and my sister are), I couldn’t imagine anything more difficult than trying to sort out the sheep from the goats (something else I have some experience with) when it came to the facts of the matter as explained by hundreds, if not thousands of folks who don’t share our obsession with binary and logical exactitude, and who were trying to describe their interaction with voting machines. She asked me if I’d heard about the Royal Mail story. I had not. But now I have.
And so have you.
Crimenutely, folks. I’m a firm believer in Hanlon’s razor. But, whether it’s malice or incompetence, it deserves to be looked at. And sometimes the inarticulate little people who are at the point of the spear have more real information than we do. This is a lesson I learned early on in my illustrious (LOL) IT career. It stood me in good stead, and I don’t think it ever, ever let me down.
Let’s hope that the truth, whatever it is (and I don’t know what the truth is–and, to be clear, neither do you) doesn’t take 15 years to come out because some of us are too cowardly, too invested in the status quo, or just too lazy, to engage with it.
Computers can, indeed, be wrong. We can talk more about whose fault that is, some other time.
Further reading:
https://en.wikipedia.org/wiki/Horizon_(IT_system)
https://www.ft.com/content/0138cd7d-9673-436b-86a1-33704b29eb60
https://www.bbc.co.uk/news/uk-england-50747143
https://www.bbc.com/news/uk-23233573
https://www.bbc.co.uk/news/uk-52905378
https://www.bbc.com/news/business-54384427
https://www.lawgazette.co.uk/news/spending-nears-40m-in-mammoth-post-office-case/5101919.article
https://www.thetimes.co.uk/article/fraud-case-costs-post-office-58m-h2xb67lgr
Published in General
If you do not investigate and harshly punish these kind of crime, they will occur again. This means individuals who acted with malice need punishment. Make your system as simple and transparent as possible.
These kind of principles apply in incident investigation for the nuclear and chemical industries. If they applied to critical IT, we might not have these kind of failures.
That was a fascinating, and somewhat horrifying, read @she. Thanks for sharing. The human costs are rarely quantified or acknowledge.
Mark Steyn once wrote about a neighbor in his small NH hamlet who caught the attention of the Fed. Ten years later, when the case finally went to trial, it took the jury two hours to find him not guilty. Yahoo! What’s missing from the story was the bankrupt business, the failed marriage, the lost home, and the unsuccessful suicide attempt.
Monty Python knew what they were talking about:
‘Mortuary Hour’
(Cut to a mortuary. Various trolleys lie about with corpses covered by sheets. Two workers are sitting at a low make-shift table with cups of tea and a transistor radio, shelling eggs and dropping them in a pickling jar.)
1st Radio Voice: … and Premier Chou En Lai, who called it ‘a major breakthrough’. Twelve men were accidentally hanged at Whitby Assizes this afternoon whilst considering their verdict. This is one of the worst miscarriages of justice in Britain since Tuesday. (music)
Yes, sadly that’s true. And not only in Britain.
From my own IT perch, I can easily see how difficult it can be to argue with someone “farther up the chain” (although that doesn’t mean superior, but in terms of supplier and customer it may always easier for the supplier to claim that the customer somehow doesn’t understand and did something wrong).
In my own situation though, I wasn’t just “consuming” someone else’s IT products, I also produced IT products myself. And I knew what I was talking about. So even if I ended up having to deal with a secretary who understood nothing, because the other actual IT people were behind layers of secretaries etc perhaps with good intentions to protect them from people who really DIDN’T know what they were talking about, I managed to get my messages through. That might be different now though, when so many more people are using computers still often without understanding the slightest thing about how they actually work.
One example that comes to mind, I was developing software for business applications – real business stuff, is what I tell people, not 3d games or spreadsheets or web sites – that included some time-keeping features.
After one update to the language compiler we all used, I found that something had changed, and it screwed up all of the programs that had previously been compiled. A 3-character data field that identified a user’s terminal, in the form of Txx where xx was their terminal number starting with 00 and going up, was appearing as 6 characters instead: Txx plus 3 spaces. (For old-timers, think of a data card where a 3-character field is being entered by the keypunch operators as 6 characters, and throwing off the position of everything after it.)
In this particular programming language, that would be seen as an issue of “working length” versus “declared length.”
When I attempted to report this to the compiler developers, I ran into a secretary, of course. But since I was an IT person myself, I didn’t just say “this isn’t working right!”
I said “Tell the compiler people that as of the last update, TERM$ is behaving as if it still has a working length of 3, but a declared length of 6.”
Anyone who knew that programming language would understand what that meant, and the implications. And, sure enough, there was a correction issued shortly thereafter.
And fortunately, nobody – especially me! – went to jail over it.
Some of my work has been with flight control systems. It is considered bad form to turn down an offer by a test pilot to accompany him when the software has been updated. You’d better believe that I tested that code until it was bulletproof.
Have you ever read the novel Hemingway’s Chair, by Michael Palin? (Yes, that Michael Palin) It is, among other things, a clever and rather pointed teardown of the British mail ethos. It was written in the 90s, when the Postal Service was busily laying the groundwork for what this damnable computer system claimed to rectify – shutting down scores of long-standing post offices, while modernizing what remained in dehumanizing ways – in short creating the later “underserved” problem by deliberately underserving people in the first place. Palin tells all this through the eyes of one of the displaced post agents.
Software crashes => aircraft crashes.
No I haven’t read it, but will investigate. In the meantime, I’m reminded of the same sort of (decades earlier) withdrawal of “service” from British Rail. That which was immortalized by Flanders and Swann:
Yeah. And I was in healthcare IT, where, on occasion, software crashes also equaled deaths. One of the reasons I always responded to the infernal pager, no matter when it went off.
Superb post! Entirely believable, and uncomfortably familiar to those of us who, if we haven’t actually participated in the mysterious chaos of subtly defective software, are at least knowledgeable of the more horrific examples — and this qualifies as one.
It amuses me that you chose to post this today, as I was having a conversation this morning about one of the problems widespread automation brings to institutions, that of inflexibility. My youngest son was born in Vietnam; when we adopted him and brought him home to America, an immigration official transposed the month and day of his birth on the immigration forms. The error went unnoticed for twenty years — until it suddenly become an almost unsurmountable wall of documentary inconsistency that made the issuance of his first U.S. passport a three year ordeal resolved with the help of a Freedom of Information Act request, the intervention of my Congresswoman’s office (that of the charming Elise Stefanik), and multiple trips to a passport office in a neighboring state.
Big, integrated, automated systems often balk at exceptions. When you encounter, not the personal authority of a knowledgeable bureaucrat, but a software system, even an infallible software system, and then discover that you’re unable to check all of the necessary boxes because your case differs just slightly from every exceptional instance anticipated by that system’s designer….
Anyway, we got his passport, he went back to Vietnam and had a nice visit, and made it home two weeks before the President stopped all travel from China and things got weird.
Many will remember all the concerns about the 1999 to 2000 date rollover…elevators would stop in their shafts, airplanes would fall out of the sky, etc…and if they didn’t fall out of the sky, the entire air traffic control system would fail and midair collisions might become common.
The woman who was running the FAA at the time celebrated New Years Eve on a commercial airline flight, together with key members of her staff.
That’s leadership.
Yeah. I spent the entire evening of December 31, 1999, in the 3E electrical closet with my employee Tony L, on the phone with the vendor, and trying to resuscitate, the only system in the entire hospital which crashed for the occasion. It was the neonatal monitoring system; that is, the system which kept track of the about-to-be-born-but-still-in-their-mothers’-womb. Important, I thought.
I remember thinking, as Tony and I looked out the very small window, at the spluttering tops of the fireworks display in Pittsburgh, about 40 miles away in real time, that I was displaying some form of leadership myself.
I’ve said before and I’ll say again now, to the extent that Y2K turned it to be a “nothing-burger,” much of it was because a lot of pre-emptive work got done once “the alarm was raised.”
I was still an undergraduate when I heard the alarm raised. That was around 1980. New software was already being designed to use four digit years. Existing systems that couldn’t be adapted because management types are none-too-bright were noted for potentially having problems later on. When they finally caught up and “not today” became “ohmigawd what are we gonna do?” the folks that fix things fixed things.
At the risk of sounding bloodthirsty, the wrong people were contemplating suicide.
Some of the software at the company I was with at that time, in the early 80s – but I wasn’t involved with that software – involved mortgage calculations which had just begun to routinely go 20 years. So that came up first.
But I actually started my work on that for the other business systems in the mid-80s. That arose from one of the reasons I went into computers to start with: I really don’t like repetitious jobs, whether that’s flipping burgers or something far more esoteric but still repetitive. With computers – at least in theory – I write a program ONCE, and then it’s the computer doing the same thing over and over, as God intended.
What I ran into though was the… repetitive… need to tweak some existing software package or another, for particular odd requirements of some customer. And that also meant having to keep multiple slightly-different versions of the software on file, and if any change or bug or whatever came up – for the payroll software it might be just some new wrinkle in tax calculation – it had to be done for all of the slightly-different versions.
So after a few “Okay, we just finished this accounting system. Next up: An accounting system!” iterations, I decided I wasn’t going to mess with that junk any more. (Especially since much of the software that would get slightly tweaked wasn’t much good to start with, no doubt in large part because it hadn’t been originally written by me!)
I determined to do a re-write of all of it, from top to bottom, and make all the little things that got tweaked one way or another for each customer, into configurable options at setup. That would mean keeping just one version of the software for all customers.
And I started with a “central calendar system” which would deal with accounting periods, holidays, appointment scheduling for the medical office billing system… The “central calendar system” was, as the name suggests, the center of it all. And it used 4-digit years.
From there and spreading out to all the other parts of the software – Accounts Payable, General Ledger, etc – dates were store internally as YYYYMMDD which allowed for easy sorting, and the display/print/input formats were determined by the selection made at installation: MM/DD/YYYY, or DD/MM/YYYY, or YYYY/MM/DD.
The other configurable options were things like type of accounting year: 12 accounting periods, or 12 regular periods plus a “reconciliation” period…
I was in the air from Florida to Rhode Island as the date rolled over. The plane was almost empty – one of the most comfortable flights I’ve ever been on. I am not easily deterred.
Yeah, I was working for Raytheon at the time and there were IT “war rooms” in facilities across the country for at least the year before, maybe longer.
Once again, I was almost 20 years ahead of the times. :-)
The only problem I encountered was before 2000. That was the year my credit card was due to expire. The expiration date came out to be “19100.” I knew right away what had happened. Someone was using a four digit year to do internal computations, but when it came time to print it out, they only needed the last two digits, so they ran X mod 1900 on the result, and used that. That would give ’88’ for 1988′, ’95’ for ‘1995’, etc. When the format for the output was calculated, they prepended ’19’ to the result. That part hadn’t been updated yet. 2000 mod 1900 = 100, so when that date was formatted, it came out ‘19100.’
Computers designed to do nothing more than count votes are never wrong. The programming is fundamentally easy. If there are errors in the count, they are not innocent.
Especially when you see fractional counts.
Important if true, but there is still a question of whether we’ve actually seen fractional counts.
A procedure-driven bureaucracy is basically equivalent to an computer-based algorithm. See my post The Reductio ad Absurdum of Bureaucratic Liberalism.
Pedant mode on:
Pretty sure that wasn’t a modulus problem. Just the normal “offset 1900” with a fixed printing prefix.
Pedant mode off:
Carry on!
You need to post your stuff here as well.
It would have said “1900” instead of “2000” then. That extra “1” came from somewhere. You could be right, though. I was just trying to figure out that “1.”
The output was formatted to put 19 before what they assumed would always be a two-digit number. When that became a 3-digit number, is when they had the problem.
Now, if they had limited what follows to be 2 digits, it would have shown as 1900.
I thought about it for a little while and liked your conclusion. (And how often do we get to say “modulus” on Ricochet, anyway?) Of course, (year – 1900) would also yield 100, and do it without division, so maybe that’s what they did and maybe that’s what Phil had in mind.
In 1980 I was working for a little software company that, among other things, created banking software running on the IBM System34. So I was doing a lot of RPG-II — something I’ve managed to forget in the subsequent 40 years. Because storage was gold back then, we scrunched dates down to save space. We had a little library we’d written to store dates as tightly as we reasonably could without making them miserable to work with in RPG-II (which wasn’t long on bit operators but did a heck of a job cranking out reports on 132 column line printers). I think we coded the century in the sign field of a zoned decimal date value, but it’s been too long and I don’t remember.
Now disk is free, so I store all my dates in Roman numerals, as a millisecond offset from a start date somewhere in the Upper Cretaceous (which choice was, I admit, a little arbitrary).