Previous research shows that,
despite its popularity, Facebook is less frequently used to share
academic content. In order to investigate this discrepancy we set out to
explore engagement numbers through their Graph API by querying the
Facebook API with multiple URLs for a random set of 103,539 articles
from the Web of Science. We identified two major challenge areas:
mapping articles to URLs and the mapping URLs to objects inside
Facebook. We then explored three problem cases within our dataset: (1)
identifying a landing page for any given URL, (2) instances where
equivalent URLs are mapped to different Facebook objects, and (3)
instances of different articles being mapped onto the same Facebook
object. We found that the engagement numbers for 11.8% of all articles
that have been shared on Facebook at least once are not reliable because
of these problems. Moreover, we were unable to identify the URL for
11.6% of the articles in our data. Taken together, the three problem
cases constitute 12.3% of the 103,539 tested articles for which
engagement numbers cannot be relied upon. Given that we only tested a
small number of problem cases and URL variants, our results point to
large challenges facing those wishing to collect Facebook metrics
programatically through the available API.