The cognitive penalties can be particularly severe for students. In one revealing experiment, researchers at Cornell University divided a class into two groups. One group was allowed to use their laptop computers to surf the web during a lecture. The other group attended the same lecture but had to keep their computers closed.
Immediately afterward, the students took a test measuring how well they remembered the lecture's content. The students who used their laptops performed significantly worse on the exam. It didn't matter, moreover, whether they surfed sites related to the subject of the lecture or unrelated sites. All the surfers performed relatively poorly.
The "daily-learning" metric for student performance that the article suggests is not the metric of performance over which students optimize. Students care about end-of-year performance, not daily learning. Students may find it optimal to learn at a time different than during class. In this case, those with computers may be maximizing their total learning over time by doing something else during class--by no means does this imply that they will have learned less at the end of the year than their counterfactual classmates who are forced to listen to an entire lecture. If there are increasing returns to studying later in the year, then students will optimally postpone their studying. If they are constrained to study more earlier in the year, then they may preform worse overall. For example, if a farmer is forced to pick low-hanging fruit from a tree early in the season, picking fruit is costly, and fruit picked early begins to decay, then that farmer will have less revenue from selling his fruit than a framer who optimizes his fruit-picking without constraints. Student performance should be measured by final output rather than by an intermediate step in a dynamic process.