AI agent for calculating screening variables in collusion detection

Collusion in procurement is an unlawful practice in which competing firms secretly coordinate their bid prices, undermining market competition. A persistent challenge for contracting authorities is the difficulty in detecting non-competitive bids, which frequently leads to the awarding of contracts at inflated prices. Screening Variables are specific indices derived from the distribution of bid values in each auction (i.e., the prices offered by bidders). These indices facilitate the processing of auction data by machine learning algorithms, enhancing their capacity to identify collusive behavior. Screening variables are instrumental not only in flagging potential collusion within individual auctions but also in detecting sustained collusive patterns among specific bidders. Typically, these variables consist of statistical indices computed directly from bid values in each procurement process. Although their computation is generally straightforward, a proper understanding of the underlying terminology is required, particularly for public officials involved in decision-making during the contract award process. This study presents the development of an artificial intelligence agent, integrated into a chatbot, designed to assist in the calculation of screening variables, thereby supporting more effective detection of collusive practices.