A Lab Experiment Using a Natural Language Interface to Extract Information from Data: The NLIDB Game

Abstract

This paper makes a case for the challenge of an inductive approach to research in economics and management science focused on the use of a natural language interface for action-based applications tailored to business-specific functions. Natural language is a highly dynamical and dialectical process drawing on human cognition and, reflexively, on economic behaviour. The use of natural language is ubiquitous to human interaction and, among others, permeates every facet of companies’ decision-making. Therefore, we take up this challenge by designing and conducting a lab experiment – conceived and named by us as NLIDB game – based on an inductive method using a novel natural language user interface to database (NLIDB) query application system. This interface has been designed and developed by us in order both (i) to enable managers or practitioners to make complex queries as well as ease their decision-making process in certain business areas, and thus (ii) to be used by experimental economists exploring the role of managers and business professionals. The long-term goal is to look for patterns in the experimental data, working to develop a possible research hypothesis that might explain them. Our preliminary findings suggest that experimental subjects are able to use this novel interface more effectively with respect to the more commons graphical interfaces company-wide. Most importantly, subjects make use of cognitive heuristics during the treatments, achieving pragmatic and satisficing rather than theoretically oriented optimal solutions, especially with incomplete or imperfect information or limited computation capabilities. Furthermore, the implementation of our NLIDB roughly translates into savings of transaction costs, because managers can make queries without recurring to technical support, thus reducing both the time needed to have effective results from business decisions and operating practices, and the costs associated with each outcome.