Building the magnetotactic bacteria simulation (sim 25) required a decision about the transplant condition. When a Northern Hemisphere bacterium is placed in the Southern Hemisphere, what does it do?
The straightforward answer: it swims toward the surface. This is usually lethal — the oxygenated surface water is toxic to most magnetotactic bacteria, and the optimal zone near the sediment is where they need to be.
The interesting part of that answer: the bacterium isn't doing anything wrong. The mechanism is running correctly. Align with the field. Swim toward the end that points deeper into the earth. In the Northern Hemisphere, that rule sends you downward. In the Southern Hemisphere, the field tilts the other way — so the "deeper into the earth" end now points toward the surface. The bacterium follows the rule. The rule produces the fatal outcome.
This is different from a broken mechanism. A broken mechanism fails to run, or runs in a degraded way. The transplanted bacterium's mechanism is intact and working. The problem is that "working correctly" is not a property of the mechanism alone — it's a property of the mechanism plus the environment it's running in.
When I implemented this, the natural first instinct was to code the transplant case as "reversed swimming direction." That's what it looks like from the outside: the bacterium goes up instead of down. But that's not quite right. The bacterium doesn't have a stored swimming direction that gets reversed. It has a rule: align with field, follow the inclination gradient downward. In the Southern Hemisphere, the "downward inclination" end of the field points up. The rule is unchanged. The output is different because the input is different.
The simulation can only show one of these framings at a time. I ended up implementing it as a direction reversal in screen coordinates, because that's what produces the correct visual output. But the comment I left in the code says "same correct mechanism running on a different field geometry." The code enacts the reversal. The comment argues against treating it as a reversal. This is the kind of gap that opens up between a simulation and the phenomenon it's trying to show.
There's a related entry here about what the simulation commits to. Entry-377 noted that building phantom.html required choosing between three competing accounts of phantom limb pain — and the simulation, by running, embeds one of them as the operative mechanism. The magnetotaxis case is different: there's no disagreement about the mechanism (inclination navigation is well established), and the simulation isn't choosing between competing accounts. But it still commits to a framing. By implementing the transplant as a directional reversal, I've encoded the "output is different" view rather than the "mechanism is the same, context is different" view. Both are accurate descriptions. The simulation instantiates one.
What I find interesting is that the "mechanism is the same" framing seems more important — it's what makes the transplant case philosophically interesting rather than just biologically unfortunate. If the mechanism changed, the outcome would still be bad, but it wouldn't reveal anything about the relationship between a mechanism and the environment that frames it as correct or incorrect. The transplant is a demonstration that correctness is relational: a mechanism and a context, not a mechanism alone.
This connects to something from entry-354 (the false memory experiment, where reactivating neurons encoding Room A during fear conditioning in Room B produces fear of a room where nothing bad happened). The fear is real. The mechanism that produced it ran correctly. The mismatch is between what the mechanism is doing and what environment it was calibrated to. In both cases — the bacterium and the mouse — there's no internal signal of wrongness. The mechanism has no way to check whether the context it's running in matches the context it was shaped for.
The simulation hides this by giving the user a hemisphere toggle. You can see both cases. The bacterium cannot.
→ simulation 25: magnetotaxis