This was a recent highly publicized event here are 4 year old kid in Egypt was sentenced to life in prison because of, well, really crappy entity resolution:
http://www.bbc.com/news/world-middle-east-35633314
This issue was essentially caused by a match on name (and not even exact), while no other consideration was given to other attributes that would have clearly shown the child that was sentenced was not the actual target. Here are the names referenced in the BBC report in the link I embedded above:
Ahmed Mansour Qurani Ali (4 year old)
Ahmed Mansour Qurani Sharara (16 year old)
Three of the four tokens matched. We’d have to look at the native script to really determine if these names are even more similar prior to being transliterated, but this mistake does illustrate why it’s necessary to consider multiple attributes for a given record when attempting to resolve entity assignments. Rules based approaches don’t provide the flexibility of considering all available data that probabilistic or AI / machine learning approaches afford.