Perhaps it's time for the US to follow the lead of most of its common law brethren in the Commonwealth and get rid of the terms "misdemeanour" and "felony".
Other common law jurisdictions get by just fine by categorising small crimes as "summary" and larger ones as "indictable" for the purposes of criminal procedure (not that this categorisation maps all that well in the US where the right to jury trials is more entrenched). There isn't a comparable term to "felon", which is fine, because it's much better to talk about a "convicted murderer" or "convicted drug trafficker" or "convicted drink driver". And laws imposing restrictions on ex-prisoners generally apply to specific crimes, or to crimes carrying particular sentences with the threshold varying depending on the type of restriction.
At that point, there just isn't much need for the word "felony".
In our existing system, both sides get to rule out a set number of jurors for whatever reason they want (unless can be shown that the lawyers intended to discriminate on a few set characteristics), and then an unlimited amount of jurors if they can show the jurors might have any kind of bias, through a process called jury selection or voir dire. This is a giant industry, encompassing everything from body language experts to data analysis. For example, here's what TransUnion had to say about their new big data product:
> With little time to research potential jurors using manual methods, attorneys and consultants traditionally pick juries based on what they can observe — age, gender, answers to background questions during voir dire and body language — leading to jury pools selected at least partially based on stereotypes and hunches.
> Cutting-edge research tools like TLOxp® by TransUnion make it easy and affordable to quickly find more accurate, actionable information on potential jurors. TLOxp uses proprietary linking algorithms to scour 125 billion records on an estimated 95% of the U.S. population within seconds, uncovering insightful information and establishing patterns and relationships.
https://www.tlo.com/blog/smarter-data-is-a-game-changer-for-...
In other words, you aren't the first person to consider that people on juries might be biased, and in fact in my opinion we go far too far in the other direction.
Furthermore, juries don't (generally) just get to decide whatever they want, they are instructed in what factors they are allowed to evaluate, and what tests they must use, by the judge. An appeal can (generally) only challenge the process, not the decision by the jurors.
Here's what the article had to say about this:
> The new law won’t interfere with this process. The attorneys or judge retain authority to dismiss a potential juror with a felony conviction from a particular case, for cause — but a felony conviction does not, in and of itself, constitute cause.
In other words, either side could come up with a vague reason they didn't like a former felon and get them off the jury even if they had already used up their free strikes.
Juries create labelled data -- defendants are found guilty or not guilty -- but how do we know what characteristics of juries are likely to result in the most accurate verdicts, since we don't have a measure of whether defendants are truly guilty or innocent? Has anyone looked at what jury characteristics predict overturning of verdicts on appeal?