A data scientist models user session durations with an exponential distribution. Sessions that are still ongoing are excluded from the dataset. What bias does this introduce?