Wednesday, 16 November 2016

Court decisions and arguments for them

Here is a little puzzle, inspired by a brief visit to the UK Supreme Court yesterday.

Suppose that an appeal, which turns purely on questions of law rather than on questions of fact, comes to a supreme court. (That is important - we are to assume that this hearing is the end of the line, so if something goes wrong, there is no further appeal procedure to put it right.) And suppose that 9 justices hear the appeal, and decide who wins by a simple majority.

5 justices would allow the appeal, and 4 would dismiss it, so the appeal succeeds. But the 5 are split 3-2. 3 of them reach their conclusion by using argument-3, and make clear their view that another argument, argument-2, would be not merely inadequate but mistaken. The other 2 rely on argument-2, and make clear their view that argument-3 would be mistaken.

The 4 who would dismiss the appeal all rely on exactly the same argument, argument-4.

Given that we expect legal decisions to be made for good reasons, should we worry that the court's decision is opposite to the decision which would be supported by the argument that attracts more support than any other argument?

Monday, 26 September 2016

Tax and software

Here is the response I have just sent to the HM Revenue & Customs consultation paper "Making Tax Digital: Bringing business tax into the digital age", which is available here:

I write in response to the consultation paper Making Tax Digital: Bringing business tax into the digital age, published 15 August 2016. I write as a private individual, not on behalf of any business or organization. (I used to be in charge of tax policy for the Institute of Directors, but I left that post in 2013.)

I only have comments on one topic, the supply of software to taxpayers to enable them to work within the proposed digital system.

The proposal is to make people use software from suppliers who have obtained the approval of HMRC, coupled with the proposal to require software suppliers to agree to provide cut-down versions (which will lack full functionality) for free.

I think this proposal addresses the problem in the wrong way, and in a way that will cause considerable trouble for HMRC. My reasons for saying this are as follows.

1. In order to comply with their legal obligations, many businesses will have to spend money in addition to paying their tax. This is fundamentally wrong. You should be able to have all necessary dealings with the state at no additional cost. (Compare the fact that if you want to go to court without spending money on a lawyer, you can do so by representing yourself.)

2. The software companies will want to offer as little as possible in the free versions, while HMRC will want a reasonable amount to be offered. HMRC will get involved in tangled and wholly unnecessary negotiations, and there will inevitably be the suspicion of cosy deals being done - unjustified suspicion, perhaps, but there nonetheless. HMRC will also be at risk of litigation from rival software providers, when they are seen as having done deals with some providers that are more generous to the providers than the deals they have done with others.

3. The negotiations with providers will go on year after year, over which updates will be provided free of charge.

It would be far better if HMRC did not get into the business of doing deals, but instead did at least one, and perhaps both, of the following. (Of the two, I think the second one, provide a full package, would be the better one.)

A. Publish the interface and allow anyone to produce software

Publish the interface - the exact requirements for any file submitted to HMRC. Then anyone could produce software to match, and anyone could make it open-source if they wanted.

It is not clear how far HMRC already intend to go down this route. Paragraph
2.14 speaks of "developing and releasing new APIs". For this to be satisfactory, it must mean release in full to the public at large (and not just authorized suppliers), and in a form that allows anyone to try out software with them. On this last point, see the remarks by Toby Parkins in response to Q250 of the evidence that the Treasury Committee took on 06 September 2016, available here:

HMRC might object that they need to authorize the software for reasons of security of the HMRC system - they don't want people finding ways to feed false data into the system. While that is a legitimate concern, I very much doubt that the authorization of software would be the remedy. After all, people will find ways to spoof authorized software, and whatever security systems will be used with the authorized software (eg public and private keys) could be made a requirement of the software that anyone was allowed to develop.

Paragraph 2.18 also indicates that HMRC would undertake lots of testing of software, something which would require HMRC only to have to test a few packages. While that might be reassuring to taxpayers, it is unlikely to be necessary. If a product is a duff one, word will quickly spread. And it would be open to HMRC and others to provide sets of test data which people could put into software, and copies of the files which should come out at the other end for sending to HMRC. Consumer groups could then run the test data and compare the files which emerged with the files which should emerge, to see whether the software was any good.

B. Provide a full package free of charge

HMRC could provide their own full-featured package to do everything, including the keeping of accounting records that would feed into submissions to HMRC, free of charge (and preferably open-source so people could make their own modified versions it if they wished, or at least look at it in detail and propose amendments), and fully functional on all of Windows, Mac and Linux.

That approach would be far more likely to wean people off the shoebox-full-of-invoices system of record-keeping than asking taxpayers to spend money on commercial systems. And it would probably involve a good deal less effort for HMRC in total than entering into negotiations with commercial suppliers and testing all their various products.

HMRC's traditional argument against providing full-featured packages free of charge is that it would undermine the private sector software industry. Well, tough. It is not HMRC's job to create work that people then have to pay to get done. The concern in paragraph 2.22 that HMRC should provide reassurance to companies that they could provide both free and paid-for packages and still make money is wholly misplaced. Companies' profitability is not in principle HMRC's problem, and it is only made HMRC's problem if they first takes the decision to rely on commercial suppliers, rather than providing a full and free package.

Moreover, acting to keep commercial providers in business right against the idea of reducing regulatory burdens. And in the old, paper records, days, the Inland Revenue happily provided all the forms you needed, free of charge.

Finally, here is an admittedly off-the-wall suggestion. If HMRC do go down this route, it will be worth considering ways to get a package developed cheaply. It would make a fine project for computer science students - some to write the package, and others to find out what was wrong with it and propose corrections. Just look at what the open-source community has achieved, using collaboration tools like GitHub, and you will see that it really would not be that difficult. HMRC's main concern would probably be time discipline - the package would have to be ready by a certain date in order to fit in with broader plans. But it is not clear that such worries would be any greater than they would be with more traditional methods of software development that have been tried by Government agencies in the past.

Monday, 25 July 2016

Pokémon Go: Philosophy Examination

Philosophy Examination

Subject: Pokémon Go

Answer as many questions as take your fancy

Time allowed: for as long as you have nothing better to do

Note: this examination is based on a simplified version of Pokémon Go, which is not identical to the actual game. The broad idea of capturing different monsters which appear on the screen of a mobile phone when one is in the right area is carried over from the actual game. Other features which are to be assumed are indicated by the questions. Candidates may assume such other features as are reasonable given what the questions ask, so long as no assumptions which would make the questions trivial to answer are made.

1. Assume that when several people are at the same location, surrounding a table, they can all see a monster in the middle of that table, and they see it in ways that reflect their relative positions. (For example, if someone facing south sees the monster face-on, someone facing west sees its right hand side.) And when a viewer moves round the table, his or her view of the monster changes accordingly.

(a) Under what additional conditions, if any, would this be evidence that the monster was real?

(b) What kind of evidence, if any, would help to settle the question of whether the viewers saw one monster or numerically distinct but qualitatively identical monsters?

(c) How would your answer to (b) be affected by whether the monster disappeared from all screens when it was captured by one of the viewers?

2. Assume that a monster is at an absolutely fixed position relative to the Earth's surface. Does it exist when no-one is looking at it?

3. What conditions, if any, of numerical identity over time could be applied to monsters?

4. New monsters are supposed to come out of eggs which have been fertilized and laid following an encounter between two monsters, and the species of a new monster is systematically related to the mother's ancestry. (It seems that these facts are known from other Pokémon games, rather than from Go.) But it appears that no-one has seen a monster lay an egg, and the eggs are simply found.

(a) Is inference to the breeding-and-laying account an example of inference to the best explanation?

(b) Is it a respectable inference (of whatever type it may be)?

5. Monsters are apparently distributed in a way that is positively correlated with human population density. Such a distribution would give people in rural areas very limited access to monsters, unless they travelled to urban areas.

(a) Assuming that there is a cost to the creation and maintenance of monsters, which is a cost per monster unrelated to the location of the monster or the number of other monsters near it, would such a distribution policy accord with utilitarian principles?

(b) If you were behind Rawls's veil of ignorance, would you approve of such a distribution policy?

6. Some of these questions come out of a conversation between the examiner and a (human) interlocutor who is happy to remain anonymous. Other people may have already had and published the same or similar thoughts.

(a) If thoughts which are qualitatively identical to those set out in the questions here have already been captured by others, do they count as captured by the examiner and the interlocutor, and do they get points for capturing them?

(b) Is there any identity other than qualitative identity for (i) thoughts (ii) Pokémon monsters?