Heavy doping is observed at high concentrations of impurities atoms. Their interaction leads to qualitative changes in the properties of semiconductors. This can be observed in heavily doped semiconductors containing impurities in such large concentrations N that the average distance between atoms of impurities (is proportional N^(1/3)) becomes smaller (or order) the average distance of electron or hole trapped by impurity atom. Under such conditions, a charge carrier may not be localized at any center, since it all the time is at a comparable distance from several identical impuritiy atoms. Moreover, the effects of impurities on the motion of electrons is generally small, as a large number of carriers with the sign of the charge opposite to the charge of impurity ions shield (ie significantly weaken) the electric field of these ions. As a result, the charge carriers introduced with these impurities are free even at very low temperatures.
Experimentally you can measure the temperature dependence of the conductivity in a wide temperature range. If there are only two characteristic regions with different slopes schedule it's a heavy doped semiconductor. If there's three regions - the third portion at low temperatures corresponds to a region of ionization of impurities, then conductor is not heavily doped.
Lev, what you describe is Mott transition in doped semiconductors, and is of course an outcome of very high doping.
Moreover, even at doping concentrations lower than the transition onset, one can talk about high doping - he would refer to the case where Boltzmann approximation to Fermi Dirac statistics does not apply, and we must use the Fermi expression.
Manal, for a specific case, you would have to compute the results with and without the approximation, and establish a criterion.
I agree with Yonatan; heavy doping is somewhat arbitrary but a good criterion is when standard statistics no longer apply. The criterion for that is when the difference between Ef and either Ec or Ev is no longer