The AGN unification model describes unobscured and obscured AGN (AGN1 and AGN2) as identical sources, with their different observed properties explained solely by orientation effects; as a result, it predicts no difference in the host galaxies. As an alternative, a second scenario has been proposed in which type-2 AGN represent an earlier stage in the life of AGN characterized by dust-enshrouded host galaxies which contribute to the obscuration and higher star formation activity, at least at earlier epochs. To test this scenario we employ Herschel data at three different wavelengths (250, 350, 500 um) to study the far-IR-to-submm properties of a sample of mid-IR selected type 2 QSOs at high redshift (1.5<z<4.5) from the FLS region (Martinez-Sansigre et al. 2005). We compare their properties to a matched sample of type-1 QSOs selected in the same field. Through SED fitting we are able to disentangle AGN and star-formation activity and consequently derive FIR luminosities of the two components, as well as SFRs and dust masses. We propose a picture in which intermediate-level radio activity in the core (pc scale) of AGN is linked to the obscuration of the nucleus (perhaps via a merger) since our AGN1 have systematically lower radio luminosities than our AGN2.