• Memory ratings are provided based on the user’s memory of
experiencing the item. When a user watches a movie at the
theater and then rates it on MovieLens or MoviePilot, they
are providing a memory rating.
• Expectation ratings are provided when the user has not yet
experienced the item (and may never experience it) but is
willing to provide a rating anyway. One example is rating
housing options — the user will likely only have the means
and opportunity to truly experience one housing situation,
but may provide ratings on other house advertisements indicating
their level of appeal to the user.
In some ways, consumption ratings are the most reliable, as the
item is fresh in the user’s mind. Memory ratings are also based on
experience, so the user has a fuller set of data on which to base their
rating than in the expectation case, but their impression of the item
may not be accurately remembered or may be influenced by various
external factors [68]. Expectation ratings are the most limited, as
the user has not experienced the item and thus is rating it based
only on the information presented about it combined with whatever
prior knowledge they may have. They can still be useful, however,
particularly in high-cost domains.