Intelligent Concept Paper—Problem Tree Generator: Using AI in writing development concept papers
Implementation: 2019
Technology: Artificial intelligence (AI)
Technology service provider: Neural Mechanics Inc.
ITD’s Partner Department: East Asia Department
In line with ADB’s Operational Approaches:
Promoting digital development and innovative technologies
Applying differentiated approaches
Delivering integrated solutions
Writing a concept paper is a necessary first step to providing assistance to ADB’s Developing Member Countries (DMCs). To produce one, writers need to manually sift through various internal documents and online materials. A single concept paper can take any time from several weeks to a month to prepare, and problem trees, which show the causes and effects of an issue that ADB seeks to help resolve, on their own can take three to five days to be completed. Even then, concept papers are unlikely to be comprehensive given limitations in finding available references. Thus, ADB sought to use technology to make writing concept papers faster and more efficient to hasten the process of providing support to DMCs. This initiative tested the feasibility of using AI to automate the generation and population of problem trees based on entered search queries.
ITD engaged Neural Mechanics Inc. as a partner in developing a minimum viable product of ICP. The tool was designed to allow users to input their own problem statements or use/edit the suggestions that were auto-generated based on their entered keywords. From there, they could add causes and effects, either from the suggested results or their inputs. Multiple responses could be added for both causes and effects. Sub-causes and sub-effects could also be included if they want to provide more specific information.
These activities were made possible because a neural network model, a subset of machine learning that could identify relationships based on attributed weights, was used to identify causal event pairs from filtered sentences. The problem tree could be generated in seconds, thanks in part to the click-and-drop feature which made it easy for users to build, edit, and update the problem tree. The tool also allowed various users to collaborate on the same problem tree and document, making work more efficient. Thus, ICP could significantly cut down the time from a month to a few seconds because users no longer need to manually research information.
The sources of information for the minimum viable product were documents from ADB’s East Asia Department and World Bank, as well as Xinhua news articles. Users could identify which of these sources the web scraper would crawl through to generate results. These were first converted to a readable format to enable the designed web scraper to crawl through these documents. These were then annotated and stored in the cloud to make document searching easier for the engine.
The initiative was launched in October 2019 and was completed in April 2020.
While ICP is no longer in use, the initiative showed that it was possible to develop a platform that could be used to build problem trees. It also marked various firsts for ADB: the first homegrown AI for ADB operations, the first one developed that leveraged the wealth of knowledge from previous works of ADB within the development landscape, and the first to harness both technology and human inputs to scan through key materials to identify problems in ADB’s DMCs.
ADB-wide consultations and/or focus group discussions may be conducted to explore whether the tool can be enhanced and used in the future.