EDIT: realising adding a comment here contributes another datapoint.
This information isn't personally identifiable by itself. Even combined with your original HN comment, the information leakage cross section and associated risk is minutely small.
I understand the concern about your privacy, but in practical terms what outcome are you afraid of in this instance?
Getting less lonely inherently requires some degree of trust in and vulnerability to other human beings.
It's tragically sad when privacy fears prevent someone from living a colorful, stimulating, and deeply fulfilling life.
In case you're open to unsolicited advice: Develop the skill of Discernment. When the "privacy threat" part of your brain activates, take a moment to analyze the degree of the threat and intentionally choose your response accordingly.
I want to approximate it to 3 mile accuracy (diagonal of a box, so in one dimension, would be sqrt(3^2 /2) = 2.12 mile in lat or long). 3 mile, in geographical coordinate would be equivalent to : 3 mile / 3960 mile (Earth radius) * 180 deg / pi rad = 0.0434 degree. I divide each dimension by 0.0434, and round it to a whole number (essentially, mod 0.0434). Then, I multiply that number by 0.0434 and save that as the location.
This means that a user's location can lie anywhere inside a box with 3 mile diagonal length at that geolocation.
If you see inconsistency in my logic, please let me know.
Given N points within M meters or so, which is roughly what you get for multiple measurements of "home" with an error on the measurement, I think you could have a reasonable stab at deriving the location from an estimate of the error on the geolocation plus which circles they fell into. All in same circle makes near centre more likely than near edge.
edit: daily check in, so all the above needs is a tendency to check in from roughly the same location each day.