During the last decade, the problem of inequality in education has been studied by several researchers, especially regarding urban and rural (inaccessible) areas. Most studies conclude that funding and adoption of new technologies are necessary to reduce inequalities. However they do not propose any specific means to overcome these inequalities. Towards this direction, in this paper a specific method is proposed that contributes to the reduction of inequalities, without demanding any funding. The method is designed especially for inaccessible areas (including isolated islands), which face network connectivity issues or Internet speed becomes very slow. In particular knowledge-based video summarization is introduced, to overcome boredom, frustration, anger etc. when students try to watch online educational videos, using limited throughput channels. The main question is: which information is less important so that it can be excluded or roughly described? In this paper students play the role of human sensors, by letting recording and analyzing their watching behaviors. Then average watching patterns are extracted and used for video summarization, which is accomplished by a keyframes extraction algorithm that detects uncorrelated content. Finally, results are presented, to delimit the potential of such applications as well as to set the bases for future work.