Category: Forecasting

We have allowed ourselves to become slaves to computers, and thereby to numbers and precision. In most contexts, precise numbers are spurious or misleading. Our obsession with them often leads to blinkered thought processes.

Over the last decade, Big Data has become big business. A vast amount of time and energy is channeled into a narrow obsession with a specific class of questions and concerns – ones that can be quantified using a computer, the more numbers the better.

Here’s an alternative approach to analysis, for businesses and individuals. Start by putting the computer to one side. Start with qualitative issues and ideas, and – where they matter – approximate numbers only, in order to get a feel for the key issues and their relative importance. Only then consider whether it’s possible and worthwhile trying to get more accurate numbers, which may or may not entail greater precision. And only then consider whether computers could help and whether it’s worth the investment in time and energy, and if so how much.

Climate analysis

Analysis of climate change and carbon targets provides a good example of our enslavement to computers. Over the last year, many institutions have become obsessed with measuring their carbon footprint. In some ways, this is good news. It’s great that awareness of the climate crisis has increased so much in twelve months. The desire to analyse is encouraging, and a natural starting point for target setting is to assess one’s (individual or corporate) current carbon footprint.

Unfortunately, this good intent often leads to a fixation with the wrong question. Instead of “What matters most?” the question that is addressed is “What can my computer work out?”.

And the latter question shapes the analysis that is undertaken. Using Paris terminology, Scope 1 and Scope 2 emissions are easy to define and determine – at least in comparison with Scope 3 – so most of the effort is concentrated on them. Yet for most institutions, with notable exceptions such as power generators and cement kilns, Scope 3 emissions are likely to be more significant. They are also harder to demarcate (how far upstream or downstream should the analysis extend?) and the corresponding data, being external to the institution, is harder to obtain.

Instead of focussing effort on working out approximate values for Scope 3 emissions, the common tendency is to ignore them, or at least make them low priority. Oil companies for example continue to focus primarily on Scope 1 and 2, even though 90% of oil-related carbon emissions occur downstream, where the oil is consumed. And Scope 3 is relatively easy to estimate for an oil company; most other companies make little or no attempt to measure their Scope 3.

Admittedly, there are other factors at play, one such being that companies don’t want to have targets for things they regard as outside their control. But, without being the only ones, data availability and ‘computability’ are certainly major drivers of how companies spend their time and energy.

Following the alternative approach mentioned above, a company or individual can fairly quickly establish what matters most as regards their greenhouse gas emissions. For an oil company, it’s their Scope 3. For a builder, it’s likely to be the upstream manufacture of the materials (Scope 3 again). For a financial institution, it’s the emissions of the companies to which it lends money, a category conveniently labelled as “Scope 3 Category 15” in the Paris agreement. For a management consultancy, say, it’s likely to be its business travel (“Scope 3 Category 6”); for a government office, it’s probably employee commuting (“Scope 3 Category 7”). For a boarding establishment (e.g. a public school, nursing home or prison), it’s probably the food served on site, which in Paris features … nowhere.

Once the question “What matters most?” has been addressed, the right priorities for action can be put in place. The oil company could (should) focus on targets for reducing hydrocarbon output. The British builder can decide to use British stone rather than stone imported from Asia. The consultancy can look at cutting back on business travel. The government office can bring in an enlightened approach to teleworking. The boarding school or care home can concentrate on the food it serves.

For this to work, we need amongst other things a cultural awakening that a computer is a tool for thought, not the embodiment of it. Though numbers matter, it matters more to determine a rough answer to a major question than a precise answer to a minor one. But computers dominate our thinking and computers thrive on numerical precision. In consequence, the entire industry is obsessed with Big Data and some of the biggest questions are parked if the numerical data is unavailable, unreliable, or hard to interpret.

I hope that in 2022 we can become more attuned to the natural world. The omens are not good.

A major problem is our obsession with computer screens and the world of AI. Instead of treating them as a tool – a very useful tool, especially for crunching numbers, but still a tool – we allow them to dominate our waking moments; and accord AI a degree of kudos it doesn’t deserve. And it matters because in so doing we forget we’re a part of nature, that the natural world is responsible for us being able to send messages, see the screen, indeed to see anything at all.

We’re allowing AI to constrain our thinking. Here’s a topical example.

A Covid Christmas

The weekend before Christmas, I caught Covid. At least I’m as certain as it’s possible to be that that’s when I caught it. Most of the family we were visiting came down with it too, at the same time. And in the preceding days, my only ventures outside were a whiz round Waitrose (ten minutes) and cycling into Reading for (ironically) a Covid booster. If I’m right, the incubation period was only two and a bit days, but that’s possible according to Harvard – especially with Omicron.

Once the PCR confirmation came through, I received emails and texts from the NHS, including about the need to respond to Track and Trace. I wanted to be helpful, so had a go, but my enthusiasm was soon evaporating.

It wasn’t so much the length of the form – but why does it have to be so long? – or the related fact that it was asking a (presumably ill) person questions that could have been asked of others (e.g. what is your partner’s NHS number? I don’t know! Why don’t you email her and ask directly, given you say you’re going to email her anyway?) Nor was it even quite the fact that you had to cross your fingers and hope common sense about exceptions would apply (“people in your household will also need to self-isolate for 10 days …”; hopefully the following page, which I couldn’t see without completing this one and pressing Continue, would have said that they’re actually exempt due to having had two doses already – but I wasn’t willing to take the risk of entering their details and have them bombarded by stay-in-your-room-during-Christmas messages even though they didn’t have to). The fact that the form appeared out-of-date didn’t engender much confidence.

No, it was none of these. The biggest issue I had with the process was the unthinking formulaic rigidity of it. After I’d given up with the electronic form, I was rung the following morning – and once I was persuaded it wasn’t spam and I spoke to a person, we tried again from where I’d left off.

The phone call did not get off to a good start. I was informed that I would have to self-isolate for 10 days. This now being the Wednesday, my response was that that statement contradicted the government announcement that I could take tests on days 6 and 7 and – providing they were both negative – be released then. Oh yes, said the caller, you’re right.

Having cleared up that none of the 20-odd work categories they list apply to me, I was then asked about where I went between the 12th and 16th. Now I’m confident I caught Covid on the 17th. I wasn’t convinced it was going to help anybody to talk about a rapid and irrelevant trip to Waitrose, or cycling to have a booster (standing masked in a socially distanced queue when I got to the venue) – and I wanted to get on to the 17th – so I answered “Nowhere”. A lie, but hopefully a white one.

So next I was asked how I thought I might have caught Covid. I said I knew exactly (an exaggeration), and then said it was visiting family that weekend. You’d think (at least I did) that might be received as an invitation to explore the family visit – who? where? did I visit anywhere else that weekend? … I had my answers prepared. Instead of which, the next words I heard were “Thank you very much, that’s all” and it was over.

I can only guess that my answers didn’t fit into the boxes. The system had decided I had caught Covid between the 12th and 16th, and there was no entry for recording details about persons visited on the 17th. Computer says No; end of call.

Non-algorithmic thinking

In the 1990s, I struggled with a couple of heavy books – The Emperor’s New Mind and Shadows of the Mind, written by the physicist Roger Penrose. Penrose, a genius who co-discovered black holes amongst his numerous achievements, was convinced that human brains cannot be adequately replicated by computers and set about trying to prove it and find a “missing science of consciousness”.

Shadows of the Mind wasn’t well received, most academics thinking that Penrose was better sticking to general relativity. Penrose’s argument – involving Godel’s theorem and “non-algorithmic thinking” – is convoluted and very hard for those of us with bog standard brains to follow. Maybe his critics were right about the details not making sense. I don’t know. But I believe his general thesis, which is that computers fundamentally differ from the brains of even the most vocal AI advocates.

It’s a minority point-of-view though. A mark of how much we revere computers is that a best-selling model is called Think Pad. Personally I know myself well enough to be confident that the best thinking I do is when staring at a blank sheet of paper, or out of the window at the garden, or going for a walk. And that computer screens after a while are positively stultifying.

Maybe others have different experience. But it certainly seems that “algorithmic thinking” isn’t doing Track and Trace any favours.