Being hella lost as rationality practice

post by Rachel Shu (wearsshoes) · 2024-06-24T23:50:59.991Z · LW · GW · 0 comments

Contents

No comments

I've had the privilege of knowing where I was only in the sense of "within this quadrangle of an insufficiently detailed map, with impassable terrain not well indicated" several times. Usually by choice. More often while alone than prudence would dictate. I think they've been pretty good informal exercises in applied Bayesian reasoning.

For illustration, I'll use the conditions I faced as the first person on a particular backpacking trail since the snow melted. Much of this trail had been washed out during the winter. This was in the High Sierras in California, so the putative trail was mostly over a mix of smooth granite surfaces, talus slopes (loose rocks and boulders), and pine forest.

Trails over solid rock. These are usually marked by a series of towers of stacked rocks, called cairns. A trail in disrepair is going to have some of these still standing, and some of them toppled or even scattered. Also, there are natural piles of rock that sort of look like cairns or toppled cairns. Finding your way between them is a matter of joint probabilities and Markov chains: your confidence of being on the trail is the odds that this pile of rocks is a cairn, given the odds that the previous pile of rocks was a cairn. Sometimes you end up deciding that you are probably no longer following a real trail and backtrack a long ways.

Trails over talus slopes. Trail planners avoid building these as they're unstable and hard to mark, but sometimes it was their only option, or, if you're off trail, the only reasonable way to get to where the trail is, or might be (usually the alternative is steep cliff face). Your best bet is often picking a sensible point on the opposite side, saying "I want to be there", and ascending, descending, traversing, and scrambling through the rocks in whatever manner feels safest and least strenuous. Perhaps you get there and it turns out that it's not where you need to be. Reorient, see if you can get to the right place from where you are, double back if you have to, and try again.

Trails through forest. These are often the most clearly marked, and most likely to look like an actual trail on the ground. They're also the hardest to find again when lost, due to reduced visibility between the trees. Paths might actually just be rivulets or deer trails. A path might go into a snowbank or thicket and be hard to find on the other side. Look for signs of human activity. Spray paint, flags attached to trees, or the smooth cut of saw-blade upon wood are pretty good indicators that the path is underfoot or nearby. Yes, but in which direction does it continue? Gather all the available evidence at hand. Multiply. Test the hypothesis you judge most probable. Update. 

Transitions between terrain types. From the forest you open out onto a meadow. Grass has overgrown the trail. Somewhere - maybe on this side after skirting the edge of the meadow, maybe on the opposite side having gone straight through the meadow, there is going to be a trail continuation back into forest. Check what you know about trail design. Is that likely to be a stream crossing? Does the trail go over that high point?

Doing this kind of pathfinding exercise is mostly frontal cortex, but your gut instincts were also built for solving this sort of thing and will have a lot of input. It's a good chance to get the two halves calibrated. When you've got a bad feeling about this line of reasoning, what implicit evidence are you tracking? What new lessons can you infer about pathfinding as a result?

Keep at it. Keep track of your decisions and your anxiety levels. Watch your step. Enjoy the sense of relief that comes when you see a fallen tree, neatly chainsawed into two sections two and a half feet apart, little tufts of grass poking out from below the exposed rings, and a thin concave band of bare dirt running through that gap, stretching into the distance.

0 comments

Comments sorted by top scores.