Obscure It or "Fudge It"?

[einsum]

@joeybom

Just last week I heard of someone who would make the circles around bird and mammal observations extra large “to account for the fact that the animal moved around and could be found anywhere within a radius of several kilometres.”

love the reasoning although it made me laugh as that circle radius of what that “several” km is is highly subjective unless they measured something statistical in field for that organism . anyway, its unrealistic to assume the pin position is to be taken as certainty of observation and circle radius is extraneous as you did although I can totally understand how the UI design as it exists can fool towards such wrong assumptions.

@joeybom

In my opinion, it would be better if it was a circle instead of a square. What happens with squares is, if there is a population in one of the corners and people make several observations just outside of that corner, you’ll get four of those squares around the population and can easily pinpoint where the exact location is. I feel this would be more difficult with circles.

Omg I was thinking of it other day. "The mathematical capability of obscuring via “axis-aligned squares versus any other shapes rather for X set of datapoints” keeping aside the caveats with any such obscurations methods as whole. let me start a new forum thread for this later.

---

@joeybom

More importantly, I think there shouldn’t be a random pin in the obscured square. When exporting observations, the coordinates for that random pin are exported. It would be better if it was just blank.

I think the anchor has to exist simply to index those observations onto map instead of overly complex patching to keep both obscured squares and points separately as is. Imagine how would u represent these on map without points in first place when they are NOT the main pin you are looking at, drawing all those background squares for those obscured observations is totally messy and so is altogether discounting them from getting a place on map.

@joeybom

Now I have to manually check every exported observation to see if the coordinates are correct, or in some random place.

you can actually select geoprivacy during exports

and even if u did single export, u can simply rely on geoprivacy data in export to separate these two sets of “obscured option” vs not obscured

@rcavasin

This is a significant flaw in how the obscuration is done because it can trip up the naive user. It’s possible to determine if the location coordinates are real or not using software. It’s just not very intuitive.

it is actually the reverse, the square on site for naive user directly already indicates the obscuration is done? if u meant export data for naive users to separate these, see above pics.

---

I used to simply obscure but now my preferences align with fudging rather as I understood how wide the obscure is (and I may not need that 30km level cell privacy) and its serious limitations (if I record a marine species on coast, and it so happens that current obscure with its naive randomization without underlying geography taken into account finds a random square and it so happens unluckily that my true point is actually just a corner of that square overlapping on this coast and thus totally defeating the guarantees of obscure, rather than fudging to 1km with geography context I could have done easily ).

Personally, I say it is actually fudging that is guaranteed to protect for my home observations over the obscure (if one is not trusting anyone so far with coordinates as is the case with my home obs) and if one is carefully taking into account of geography during such fudge to find a good harder fudge circle, than relying on “obscure” - but ofc also doing fudging of their other spatiotemporal data equally.

for discussion fullness, here is an approximate graph of how wide the current obscuration cell is:

@einsum

for anyone wondering, that would be d = 0.2*111.32*sqrt(1 + cos(lat_deg)^2) and here is a plot

image908×502 40.5 KB

so that point would have gotten mapped if its atmost 30.4km accurate i guess and better stay within 20km radius if you are manually picking.