CHS: Medium: CAST: Child Adaptive Search Tool

The goal of the project is to empower emergent searchers -- initially children ages 6-11 -- by researching, designing, and developing search tools that improve their information literacy and searching capabilities through modeling and adapting to their abilities as they search. Current search engines, even ones specifically designed for children, offer weak support for children's search needs due to their developing skills related to spelling, language use (including synonyms), understanding categories, refining queries, and evaluating relevance and quality of results. This makes it hard for children to create effective queries, use the results suggested by the search engine, and understand the relationship between the queries and the results returned. Bringing together expertise in human-computer interaction, information retrieval, natural language processing, and education, the project team will both (a) further scientific understanding of children's search abilities, and (b) design tools to support it through the iterative development of CAST (Child Adaptive Search Tool), designed for children aged 6-11. CAST will be designed to model and respond to users' literacy and maturity levels as well as search intent missing from their formal queries. For example, when a child submits the query "Tiger", CAST will tend to prioritize tiger habitat or Winnie the Pooh's friend Tigger, which likely correlate better to a child's search intent than information on Tiger Woods. To reach this goal, the team will collaborate with children and teachers throughout the course of the project, working with partners in several local schools to increase the impact of the application itself and to improve the dissemination of the results. The results on supporting search in the special population of children in this research also have the potential to inform similar problems and methods aimed at other populations who might have systematic differences in their search ability, from older adults to second language speakers.The team will use two main lenses to guide the project work. The first is an educational research lens that grounds design activities in the known needs of children, for example, using spelling development research to create models of spelling correction that are tailored to this audience rather than general dictionary-based spelling correction algorithms. The second is a version of participatory design called cooperative inquiry that closely involves both children and teachers along with the research team throughout the design process, allowing the team to benefit from children's direct perceptions of their own needs and ideas for addressing them as well as the teachers' knowledge of children's needs when learning to search for information to support their own education. The work will proceed in three main phases. The first phase focuses on defining requirements and resources, through partnering with local public schools and collecting publicly available datasets of children's search behavior, vocabulary lists, popular websites for kids, and other resources that can inform models and algorithms tuned to children's search. The second phase involves designing and developing an initial version of CAST that includes methods to support query formulation, spelling, and navigating ontologies in order to help children better express their search needs, along with tools such as enhanced result snippets for evaluating the quality and relevance of results. The third phase will focus on evaluating both individual components and overall system quality; this includes using quantitative analysis of observable behaviors in the system such as queries, session lengths, and returned result quality as well as direct testing with children and teachers to evaluate their perceptions of the system's usability and their willingness to use it going forward. The team will use the evaluations to iteratively measure both individual components and the overall system, which will be deployed at a number of public libraries in the project team's home state for use by children outside of the research context.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

The goal of the project is to empower emergent searchers -- initially children ages 6-11 -- by researching, designing, and developing search tools that improve their information literacy and searching capabilities through modeling and adapting to their abilities as they search. Current search engines, even ones specifically designed for children, offer weak support for children's search needs due to their developing skills related to spelling, language use (including synonyms), understanding categories, refining queries, and evaluating relevance and quality of results. This makes it hard for children to create effective queries, use the results suggested by the search engine, and understand the relationship between the queries and the results returned. Bringing together expertise in human-computer interaction, information retrieval, natural language processing, and education, the project team will both (a) further scientific understanding of children's search abilities, and (b) design tools to support it through the iterative development of CAST (Child Adaptive Search Tool), designed for children aged 6-11. CAST will be designed to model and respond to users' literacy and maturity levels as well as search intent missing from their formal queries. For example, when a child submits the query "Tiger", CAST will tend to prioritize tiger habitat or Winnie the Pooh's friend Tigger, which likely correlate better to a child's search intent than information on Tiger Woods. To reach this goal, the team will collaborate with children and teachers throughout the course of the project, working with partners in several local schools to increase the impact of the application itself and to improve the dissemination of the results. The results on supporting search in the special population of children in this research also have the potential to inform similar problems and methods aimed at other populations who might have systematic differences in their search ability, from older adults to second language speakers.The team will use two main lenses to guide the project work. The first is an educational research lens that grounds design activities in the known needs of children, for example, using spelling development research to create models of spelling correction that are tailored to this audience rather than general dictionary-based spelling correction algorithms. The second is a version of participatory design called cooperative inquiry that closely involves both children and teachers along with the research team throughout the design process, allowing the team to benefit from children's direct perceptions of their own needs and ideas for addressing them as well as the teachers' knowledge of children's needs when learning to search for information to support their own education. The work will proceed in three main phases. The first phase focuses on defining requirements and resources, through partnering with local public schools and collecting publicly available datasets of children's search behavior, vocabulary lists, popular websites for kids, and other resources that can inform models and algorithms tuned to children's search. The second phase involves designing and developing an initial version of CAST that includes methods to support query formulation, spelling, and navigating ontologies in order to help children better express their search needs, along with tools such as enhanced result snippets for evaluating the quality and relevance of results. The third phase will focus on evaluating both individual components and overall system quality; this includes using quantitative analysis of observable behaviors in the system such as queries, session lengths, and returned result quality as well as direct testing with children and teachers to evaluate their perceptions of the system's usability and their willingness to use it going forward. The team will use the evaluations to iteratively measure both individual components and the overall system, which will be deployed at a number of public libraries in the project team's home state for use by children outside of the research context.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.