I guess it would be preferable to getting all the disembodied "x laughed at an image" or "x loved an image" messages without even knowing which image it refers to. One could simply ignore the feature if it were less intrusive.