Computer Collates Climate Contrarian Claims

Coan et al. in Nature have an interesting text analysis of climate skeptics’ claims.

I’ve been at this long enough to notice that a few perennial favorites are missing, perhaps because they date from the 90s, prior to the dataset.

The big one is “temperature isn’t rising” or “the temperature record is wrong.” This has lots of moving parts. Back in the 90s, a key idea was that satellite MSU records showed falling temperatures, implying that the surface station record was contaminated by Urban Heat Island (UHI) effects. That didn’t end well, when it turned out that the UAH code had errors and the trend reversed when they were fixed.

Later UHI made a comeback when the SurfaceStations project crowdsourced an assessment of temperature station quality. Some turned out to be pretty bad. But again, when the dust settled, it turned out that the temperature trend was bigger, not smaller, when poor sites were excluded and TOD was corrected. This shouldn’t have been a surprise, because windy day analsyses and a dozen other things already ruled out UHI, but …

I consider this a reminder of the fact that part of the credibility of mainstream climate science arises not from the fact that models are so good, but because so many alternatives have been tried, and proved so bad, only to rise again and again.

Spreadsheets Strike Again

In this BBC podcast, stand-up mathematician Matt Parker explains the latest big spreadsheet screwup: overstating European productivity growth.

There are a bunch of killers in spreadsheets, but in this case the culprit was lack of a time axis concept, making it easy to misalign times for the GDP and labor variables. The interesting thing is that a spreadsheet’s strong suite – visibility of the numbers – didn’t help. Someone should have seen 22% productivity growth and thought, “that’s bonkers” – but perhaps expectations of a COVID19 rebound short-circuited the mental reality check.

ChatGPT struggles with pandemics

I decided to try out a trickier problem on ChatGPT: epidemiology.

This is tougher, because it requires some domain knowledge about terminology as well as some math. R0 itself is a slippery concept. It appears that ChatGPT is essentially equating R0 and the transmission rate; perhaps the result would be different had I used a different concept like force of infection.

Notice how ChatGPT is partly responding to my prodding, but stubbornly refuses to give up on the idea that the transmission rate needs to be less than R0, even though the two are not comparable.

Well, we got there in the end.

ChatGPT and the Department Store Problem

Continuing with the theme, I tried the department store problem out on ChatGPT. This is a common test of stock-flow reasoning, in which participants assess the peak stock of people in a store from data on the inflow and outflow.

I posed a simplified version of the problem:

Interestingly, I had intended to have 6 people enter at 8am, but I made a typo. ChatGPT did a remarkable job of organizing my data into exactly the form I’d doodled in my notebook, but then happily integrated to wind up with -2 people in the store at the end.

This is pretty cool, but it’s interesting that ChatGPT was happy to correct the number of people in the room, without making the corresponding correction to people leaving. That makes the table inconsistent.

We got there in the end, but I think ChatGPT’s enthusiasm for reality checks may be a little weak. Overall though I’d still say this is a pretty good demonstration of stock-flow reasoning. I’d be curious how humans would perform on the same problem.

Can ChatGPT generalize Bathtub Dynamics?

Research indicates that insights about stock-flow management don’t necessarily generalize from one situation to another. People can fill their bathtubs without comprehending the federal debt or COVID prevalence.

ChatGPT struggles a bit with the climate bathtub, so I wondered if it could reason successfully about real bathtubs.

The last sentence is a little tricky, but I think ChatGPT is assuming that the drain might not be at the bottom of the tub. Overall, I’d say the AI nailed this one.

ChatGPT does the Climate Bathtub

Following up on our earlier foray into AI conversations about dynamics, I decided to follow up on ChatGPT’s understanding of bathtub dynamics. First I repeated our earlier question about climate:

This is close, but note that it’s suggesting that a decrease in emissions corresponds with a decrease in concentration. This is not necessarily true in general, due to the importance of emissions relative to removals. ChatGPT seems to recognize the issue, but fails to account for it completely in its answer. My parameter choice turned out to be a little unfortunate, because a 50% reduction in CO2 emissions is fairly close to the boundary between rising and falling CO2 concentrations in the future.

I asked again with a smaller reduction in emissions. This should have an unambiguous effect: emissions would remain above removals, so the CO2 concentration would continue to rise, but at a slower rate.

This time the answer is a little better, but it’s not clear whether “lead to a reduction in the concentration of CO2 in the atmosphere” means a reduction relative to what would have happened otherwise, or relative to today’s concentration. Interestingly, ChatGPT does get that the emissions reduction doesn’t reduce temperature directly; it just slows the rate of increase.

Modeling with ChatGPT

A couple weeks ago my wife started probing ChatGPT’s abilities. An early foray suggested that it didn’t entirely appreciate climate bathtub dynamics. She decided to start with a less controversial topic:

If there was a hole that went through the center of the moon, and I jumped in, how long would it take for me to come out the other side?

Initially, it’s spectacularly wrong. It gets the time-to-distance formula with linear acceleration right, but it has misapplied it. The answer is wrong by orders of magnitude, so it must be making a unit error or something. To us, the error is obvious. The moon is thousands of kilometers across, so how could you possibly traverse it in seconds, with only the moon’s tiny gravity to accelerate you?

At the end here, we ask for the moon’s diameter, because we started a race – I was building a Vensim model and my son was writing down the equations by hand, looking for a closed form solution and (when the integral looked ugly), repeating the calculation in Matlab. ChatGPT proved to be a very quick way to look up things like the diameter of the moon – faster even than googling up the Wikipedia page.

Since it was clear that non-constant acceleration was wrong, we tried to get it to correct. We hoped it would come up with F = m(me)*a = G*m(moon)*m(me)/R^2 and solve that.

Ahh … so the gigantic scale error is from assuming a generic 100-meter hole, rather than a hole all the way through to the other side. Also, 9.8 m/s^2 is Earth’s surface gravity.

Finally, it has arrived at the key concept needed to solve the problem: nonconstant acceleration, a = G*M(moon)/R^2 (where R varies with the jumper’s position in the hole).

Disappointingly, it crashed right at the crucial endpoint, but it’s already done most of the work to lay out the equations and collect the mass, radius and gravitational constant needed. It’s still stubbornly applying the constant acceleration formula at the end, but I must say that we were pretty impressed at this point.

In the same time, the Vensim model was nearly done, with a bit of assistance on the input numbers from Chat GPT. There were initially a few glitches, like forgetting to reverse the sign of the gravitational force at the center of the moon. But once it worked, it was easily extensible to variations in planet size, starting above or below the surface, etc. Puzzlingly the hand calculation was yielding a different answer (some kind of trivial hand computation error), but Matlab agreed with Vensim. Matlab was faster to code, but less interactive, and less safe because it didn’t permit checking units.

I’d hesitate to call this a success for the AI. It was a useful adjunct to a modeler who knew what they were doing. It was impressively fast at laying out the structure of the problem. But it was even faster at blurting out the wrong answer with an air of confidence. I would not want to fly in a plane designed by ChatGPT yet. To be fair, the system isn’t really designed to do physics, but a lot of reasoning about things like the economy or COVID requires some skills that it apparently doesn’t yet have.

Controlled Burn, Wood Stove, or Dumpster Fire?

The Twitter mess is a really interesting example of experimenting on a complex system in real time, apparently without much of a model.

I think fire is an interesting analogy (as long as you don’t take it too seriously, as with all analogies). There are different kinds of fires. A controlled burn improves forest health and lowers future risk by consuming dead wood. I think that’s what Musk is trying to accomplish. A fire in a wood stove makes nice heat, regulated by air flow. Controlled growth may be another Musk objective. An uncontrolled burn, or a burn where you don’t want it, is destructive.

I think the underlying parallel is that fire is driven by reinforcing feedback, and any organization has a lot of positive feedback loops. Success requires that the virtuous cycles are winning over the vicious cycles. If you get too many of the bad reinforcing feedbacks going, you have an uncontrolled burn in a bad place. This is often fatal, as at Sears.

Here are some of the loops I think are active at Twitter.

First, there’s the employee picture. I’ve divided them into two classes: over- and under-performing, which you might think of as identifying whether they produce more team value than their compensation indicates, or less. The dynamics I’ve drawn are somewhat incomplete, as I’ve focused on the “over-” side, omitting a number of parallel loops on the “under-” side for brevity.

There are some virtuous cycles you’d like to encourage (green). Hiring better people increases the perceived quality of colleagues, and makes it easier to recruit more good people. As you hire to increase work capacity, time pressure goes down, work quality goes up, you can book more work in the future, and use the revenue to hire more people. (This glosses over some features of IT work, like the fact that code is cumulative.)

There are also some loops you’d like to keep inactive, like the orange loop, which I’ve named for mass exodus, but might be thought of as amplifying departures due to the chaos and morale degradation from initial losses. A similar loop (not colored) is triggered when loss of high-performing employees increases the workload on the remainder, beyond their appetite.

I suspect that Musk is counting on mass layoffs (red) to selectively eliminate the underperforming employees, and perhaps whole functional areas. This might work, except that I doubt it’s achievable without side effects, either demoralizing good employees, or removing functions that actually made unobserved vital contributions. I think he’s also counting on promises of future performance to enlist high performers in a crusade. But mass layoffs work against that by destroying the credibility of promises about the future – why stick around if you may be terminated for essentially random reasons?

Another key feature (not shown) is Musk’s apparent propensity to fire people for daring to contradict him. This seems like a good way to selectively fire high performers, and destroy the morale of the rest. Once you ignite the vicious cycles in this way, it’s hard to recover, because unlike forest detritus, the overperformers are more “flammable” than the underperformers – i.e., they have better prospects at other companies. Having the good people leave first is the opposite of what you want.

How far this fire spreads depends on how it impacts customers. The initial mass layoffs and reinforcing departures seem to have had a big impact on moderation capacity. That triggers a couple more vicious cycles. With moderation capacity down, bad actors last longer on the platform, increasing moderation workload. Higher workload and lower capacity decreases quality of moderation, so the removal of bad accounts falls more (red). As this happens, other potential bad actors observe the opportunity and move into the breach (orange).

There are some aspects of this subsystem that I found difficult to deal with on a CLD. The primary questions are of “good and bad from whose perspective,” and whether greater intentional permissiveness offsets diminished moderation capacity. I think there are some legitimate arguments for permitting more latitude (“sunshine is the best remedy”) but also compelling arguments for continued proscription of certain behavior (violence for example). The credibility of policy changes so far, such as they can be determined, is undermined by the irony of the immediate crackdown on freedom to criticize the boss.

One key feature not shown here is how advertisers view all this. They’re the revenue driver after all. So far they seem to fear the increase in turbulence and controversy, even if it brings diversity and engagement.That’s bad, because it’s another vicious cycle (roughly, less revenue -> less capacity -> more conflict -> less revenue).

Account holders might become more of a revenue driver, but the initial rollout of the $8 verification idea was badly botched – presumably in part because of the simultaneous mass reduction in organizational capacity. This is unfortunate, because reducing anonymity might be a good way of promoting quality information through accountability.

The alternative, if Twitter collapses, is not entirely appetizing. Other big platforms aren’t exactly paragons of freedom or civility, and alternatives like Mastodon that provide more self-moderation capacity probably also enhance the insularity of filter bubbles.

I’m wondering again, (how) Should Systems Thinkers be on Social Media?


AI is killing us now

The danger of path-dependent information flows on the web

Encouraging Moderation