Team JSTOR/Group 1: Morgan Burton Isabela Carvalho Stan (Tze-Hsiang) Lin Leo (Lei) Shi Data for Research (DfR) for JSTOR
Introduction to DfR System that includes metadata, information visualization, and article retrieval for JSTOR articles JSTOR is a major database of scholarly articles Provides “facets” or “selectors” that allow the user to filter their search based on specific elements such as journal, author, and discipline Provides graphs that update dynamically based on search query User base: User might be a researcher such as a doctoral student in linguistics, or a more casual researcher interested in comparing trends across disciplines (not exhaustive)
Methods Interaction map Provides a map of the sections of the site Personas and Scenarios A glimpse at what the typical user and situation might be for the system 5 Interviews conducted on potential users Comparative Analysis We assessed several competitive systems including Google Scholar and NINES Survey We surveyed over 20 target users Heuristic Evaluation An evaluation of general usability principles Usability Testing of 5 target users
Chart for interaction map Diff. Views: Charts, Graph Results List Keyterm Cloud Whole Data Set Of DfR Refined Data Set Narrowing Down by USER
Finding: The overall purpose of DfR is clear to users at first glance – prior to interacting with the system Usability testing result: we tested prior finding from heuristic analysis that purpose of site might be unclear at first glance We asked users to fill out pre-task assessments where we asked them to answer questions about their expectation of the system Form asked users about what their general idea of the site was Result: User expectation matched what site was about and accurately inferred relationship to JSTOR
Evidence and recommendation Some answers provided by users: “The statistics about the publications, categorized by publication year, discipline.” “I think it's a site that gives information about articles published on certain topics.” “Searching for scholarly articles by date and discipline/area.” “This is a websites showing some statistics about paper publications and properties in JSTOR.” Recommendation: (contrary to prior finding) do not include an explanatory sentence on the main page about DfR Users have a good sense of what DfR is and what its relationship to JSTOR is 6
Finding: lack of visual indication of interrelationship between search and select features ’Results list’, ‘key term’, and ‘references profile’ features are tightly linked to the main search Current layout does not give an indication that ‘results list’, ‘key term’, and ‘reference profile’ are not separate content, but are about the search query done on the main page There is a hierarchy Evidence: usability test Some users did not understand that under the article list they would see the results of the search done on the main page
Comparison of versions Location indicates incorrect hierarchy Current Version Appearance of being in the same frame indicates closer relationship Previous Version
Older version took advantage of proximity
Recommendation Move search bar to a higher level in order to indicate the hierarch between search and given search elements: the given elements are under the search level Have the links placed under the search bar, inside the grey box.
Cognitive model & usability
Designer v. User
“It’s like Google Scholar” Instances of expectations v. reality using Data for Research
Finding: The cognitive model of users and design of DfR are divergent.
The way people think for the purposes of comprehension and prediction Significance: for people to understand how to use the Data for Research tool, designers must understand the way they already think Usability: After purpose, there must be positive interaction in function for repeat use Cognitive Model: Defined
“It’s like Google Scholar” (but it isn’t!)
Refined Data Set #1 Refined Data Set #2 Refined Data Set #1 Refined Data Set #2 Whole Data Set Of DfR Whole Data Set Search #1 Search #2 Search #2 IF NOT “Clear All” Search #1 Other Database Search Search in DfR
Instance: Search aggregation - search terms accumulate, rather than reset on new search (EXCEPT WHEN going directly to index)
Instance: Keyword searching + blank spaces - all produce DIFFERENT search results - punctuations have different treatment in the DfR interface
Recommendations Search aggregation: Clearer path for new search vs. adjusting current search (“New Search” button) Keyword punctuation: Choose & specify one punctuation as AND operator Clarify how search results are accumulated (using all terms? listing by articles and journals with higher frequency?)
Search record is crucial to researchers - must keep track of information they gather Duplication of search in results view indicates system action to users Instances After-search feedback Facets/Selectors Finding: A lack of DfR system feedback left searches with unclear meanings.
Lack of system feedback before and after making a search - No tracking or matching of search terms No indication that anything has happened! - Selection criteria box is not prominent enough to notice
Facets/Selectors New version: Not intuitive that the NAMES are links Further, cannot determine what they are doing to the results (start with selection ALL included?) Older version: Check and “X” boxes Much clearer
intuitive as to what is happening when “checking” (adding) or “X”-ing (subtracting) aspect of information
Recommendations Search Feedback: Additional feedback after search that indicates search has been performed Google Scholar model: redundancy WORKS! Header renaming to “Search Results” Facet/Selector Appearance: Reinstate the "X" function for all selectors (option to eliminate from results or from search entirely) Reinstate "checkmark" function for facets that have been eliminated or are not included in the results.
Summary For (finding 1)...for marketing purposes, a description of DfR is NOT needed on the main page - it’s intuitive to users! For (finding 2)...take advantage of X to Y. <-- not sure what to put here. For (finding 3)...similar cognitive models will lead to positive interactions between the system and new users. For (finding 4)...clear feedback leads to discernible meaning of search results.