The data quality assessment is a summary of an observation's accuracy. All observations start as "casual" grade, and become "needs ID" when

• the observation has a date
• the observation is georeferenced (i.e. has lat/lon coordinates)
• the observation has photos or sounds
• the observation isn't of a human

Observations become "research grade" when

• the iNat community agrees on species-level ID or lower, i.e. when more than 2/3 of identifiers agree on a taxon

Observations will revert to "casual" if the above conditions aren't met or the community agrees

• the location doesn't look accurate (e.g. monkeys in the middle of the ocean, hippos in office buildings, etc.)
• the organism isn't wild/naturalized (e.g. captive or cultivated by humans or intelligent space aliens)
• the observation doesn't present evidence of an organism, e.g. images of landscapes, water features, rocks, etc.
• the observation doesn't present recent (~100 years) evidence of the organism (e.g. fossils, but tracks, scat, and dead leaves are ok)
• the observation no longer needs an ID and the community ID is above family
• the observer has opted out of the community ID and the community ID taxon is not an ancestor or descendant of the taxon associated with the observer's ID

And if that wasn't complicated enough, there are also situations where the system gets a vote:

• The system will vote that the observation is not wild/naturalized if there are at least 10 other observations of a genus or lower in the smallest county-, state-, or country-equivalent place that contains this observation and 80% or more of those observations have been marked as not wild/naturalized.