1. Run an opt-out SHIELD study to answer the question: "how many people can find an 'opt-out' button?". That's all. You launch this at people with as much notice as you would plan on doing for RAPPOR, and see if you get a 100% response rate. If you do not, then 100% - whatever you get are going to be collateral damage should you launch DP as opt-out, and you need to own up to saying "well !@#$ them".
2. Implement RAPPOR and then do it OPT-IN. Run three levels of telemetry: (i) default: none, (ii) opt-in: RAPPOR, (iii) opt-in: full reports. Make people want to contribute, rather than trying to yank what they (quite clearly) feel is theirs to keep. Explain how their contribution helps, and that opting-in could be a great non-financial way to contribute. If you give a shit about privacy, work the carrot rather than the stick.
3. Name some technical experts you have consulted. Like, on anything about DP. The tweet stream your intern sent out had several historical and technical errors, and it would scare the shit out of me if they were the one doing this.
4. Name the lifetime epsilon you are considering. If it is 0.1, put in plain language that failing to opt out could disadvantage anyone by 10% on any future transaction in their life.
I think the better experiment that is going on here is the trial run of "we would like to take advantage of privacy tech, but we don't know how". I think there are a lot of people who might like to help you on that (not me), and I hope you have learned about how to do it better.