Hu et al. describe a music recommendation system for drivers based on the driver’s social condition and fatigue level and the road condition. It uses a smartphone for local calculation and the cloud for heavy recommendation computation. After listening to the music recommended by the system, the fatigue level and the degree of negative moods will be lower compared to the drivers who do not use the system. Thus, the road will hopefully be safer. The system is called SAfeDJ (short for “safe disc jockey”).

First of all, the fatigue level is both an input and a measure of the system output, so there should be a continuous feedback system; the authors didn’t consider this. Second, fatigue level and moods, though related, do not entirely equate to driving safety. The test of this paper is under a controlled environment. If put to a real road situation, the results may be very different due to the many variables in a real driving situation. It will be hard to tell which factors are more important in improving driving safety. If we have sensors to detect the fatigue level, why not keep warning and reminding the drivers to take a rest instead of using suitable music to improve road safety? Playing music that the driver likes will not help much.

In the system, the most difficult part is to install and connect all those sensors. Calculation is relatively easy after we have data, but this part is left out of the paper completely. Also, Figures 1 and 2 are the most important and should be explained in detail. Figure 3 is not necessary, and the personal data is obtained from the user profiles of Facebook, a very unreliable and inaccurate source.

Finally, a driver’s personal traits and habits are the most vital to driving safety. How to detect these bad behaviors is much more important than listening to suitable music.