Condition 3 was a more traditional DG utilizing the same
population and currency and offering anonymity in a manner very
similar to that employed by Hoffman and colleagues (Hoffman et al.,
1994). Individuals waiting alone at a bus stop within a block of a
casino in Las Vegas, Nevada were approached and asked if they
wished to participate in a study that would involve them receiving
some money. If they agreed, they were then consented only to the DG
portion of the study and not the questionnaire (this was because the
participants in Conditions 1 and 2 were unaware of the $10 payment
when they made their donation decisions). In this experiment,
participants were handed two envelopes, a yellow envelope with
the word “KEEP” on it, and a white envelope with the word “GIVE” on
it. They were then given $20 worth of chips as well as six fake chips.
The chips were handed to them in a small chip-holder box with a
divider in the middle and “FAKE” written on one side and “REAL”
written on the other. While the researcher's back was turned, they
divided the real chips between the two envelopes as they desired, and
then distributed the fake chips between the two envelopes so as to
make them feel roughly the same. In this way, the researcher
remained totally unaware as to the composition of the chips in the
donation envelope. It was explained to them that the chips in the
“GIVE” envelope were to be recorded by a separate research assistant
at a later date and handed out to a random individual waiting at a bus
stop. They were to take those in the “KEEP” envelope home with them.

I think this renders your story a bit less plausible, since apparently the subject isn't having face-to-face contact with anyone who learns their choice.

(Disclaimer: I was just skimming over the paper and happened to see this bit--in the same way I made you seem wrong by finding this, someone else may make me seem wrong by reading the entire paper or the DG literature in depth. Unknown unknowns and whatnot.)