Navigators based on real-time traffic achieve suboptimal results since, in face of congestion, they greedily shift drivers to currently light-traffic roads and cause new traffic jams. This paper presents Themis, a participatory system navigating drivers in a balanced way. By analyzing time-stamped position reports and route decisions collected from the Themis application, the Themis server estimates both the current traffic rhythm and future traffic distributions. According to the estimated travel time and a popularity score computed using the learned information, Themis coordinates traffic between alternatives and proactively alleviates congestions. Themis has been implemented and its performance has been evaluated at different penetration rates based on real data. Experiments using data from 26,000 taxis demonstrate that Themis reduces both traffic congestions and average travel time at various penetration rates as low as 7%.