<p>It is challenging to achieve consensus under the constraint of a single threshold in multi-attribute large-scale group decision-making (MALSGDM) within a probabilistic linguistic environment, particularly in complex scenarios such as selecting the optimal site for a wastewater treatment plant (WWTP). The site selection process for a WWTP involves integrating the opinions of multiple decision-makers (DMs) and requires comprehensive consideration of environmental, economic, and social attributes, where decision information is often characterized by fuzziness and uncertainty. Consequently, a decision framework which integrates learning mechanism and feedback mechanism is proposed to ensure the completeness of evaluation information and maintain the consistency during the consensus reaching process (CRP). However, most existing methods rely on the constraint of a single threshold, which makes it difficult for DMs to reach consensus in a fuzzy environment, particularly for probabilistic linguistic term sets (PLTSs) information. Therefore, it is important to explore methods for ensuring that DMs reach consensus at a low range of adjustment with adaptive threshold constraints. To achieve this goal, a new learning mechanism based on PLSs is proposed for decision information processing, and a feedback mechanism with adaptive multi-threshold constraints is introduced into the consensus recognition rules. Moreover, an optimization method based on the minimum adjustment consensus model (MACM) is proposed for the CRP. Finally, a case study on selecting the optimal site for a WWTP demonstrates the effectiveness and flexibility of the proposed methods. This framework is not only applicable to WWTP’s site selection but can also be extended to other domains requiring MALSGDM.</p>