Learning to Propose Amendments: Identifying Patterns in the Right to Information Query Log

摘要

The Right to Information (RTI) Act 2005 provides Indian citizens with the platform to access information present in government offices. Every government office has the RTI query-log data containing citizens' interaction with the government. The objective of this work is to analyze the RTI query-log data to identify underlying latent patterns such that the identified latent patterns have the potential to act as feedback for amendments to laws. We, for the first time, collect RTI queries and reply-statistics from government educational institutions across India. Three latent patterns are quantified namely (i) transparency of institutions from RTI query-reply statistics (ii) factors that influence the transparency of institutions (iii) effectiveness of implementation of the RTI act. A two-dimensional institute-query-category matrix representation of the RTI data is proposed. The above three latent parameters are simultaneously quantified via maximum likelihood estimation of the data matrix using Graded Response Model. Through the extracted latent patterns we identify inconsistencies in the execution of the RTI act and indicate a feedback for potential amendment to the RTI act.