Aiming at the consensus problem of multi-attribute group decision-making in a social trust network context, where decision-makers express their preferences by means of Double Hierarchy Linguistic Term Set and both the initial trust information and evaluation information are incomplete, this study proposes a novel time-varying aggregation criterion based on subjective and objective information, and designs a dual trust-evaluation feedback mechanism to facilitate the achievement of group consensus. Firstly, an improved shortest path model and the authoritative preferences of the trust network are employed to effectively complement the missing entries in the trust and evaluation matrices. Secondly, by integrating subjective and objective information and introducing adaptive weighting based on the coefficient of variation, a time-varying aggregation model that balances trust evolution and evaluation stability is constructed to generate the group ranking vector of alternatives. Thirdly, relying on the Spearman rank correlation coefficient, a dual trust-evaluation feedback mechanism is developed to dynamically guide individuals to revise their preference expressions until group consensus is attained, thereby deriving the final decision outcomes. Finally, case applications, simulation experiments, and comparative analyses are conducted to verify that the proposed method can effectively resolve the dilemma of incomplete information, and exhibits distinct advantages in excavating potential information and enhancing the quality of group consensus.