An evaluation of computer-supported backtracking in a hierarchical database
A common concern for people using computer databases is becoming "lost" within the complex hierarchy of entries. Most direct manipulation interface design guidelines suggest designers should include a feature for “undoing” user inputs (Smith and Mosier, 1986). In the case of a database, undo translates to backtracking support. The first purpose of this research was to confirm that computer-supported backtracking tools reduce navigation time over manual backtracking. The second purpose was to compare navigation times among a subset of backtracking tools. The third purpose was to determine if users prefer to use one or more backtracking tools significantly more than others.
Four backtracking tools were developed by crossing two factors: History (history list vs no history list) and Level (component vs entry). History list indicates the user may view a chronological listing of nodes that have been viewed and directly select a destination node. No history list means the user must backtrack through each visited node with no shortcuts. Component indicates the backtracking tools operate only at the lowest level, or smallest definable node, of the tree-like database structure. Entry means that backtracking occurs at the higher parent node. Thus, multiple components make up an entry . In addition to the four computer backtracking tools, overall navigating and manual backtracking was done using a hierarchical Table of Contents.
The tools were evaluated in an experimental, hierarchical, direct-manipulation database. Trials were conducted in the form of a multiple-choice information retrieval task. The independent variables included the backtracking tool (four-computer supported, one-manual) and the backtrack Task Length. The dependent measures included navigation time, the frequency with which the computer tool was used over manual backtracking (Table of Contents), and questionnaire responses.
The results of this study provided some of the first solid support for the many guidelines that have been written recommending user recovery, or undo support. Backtracking with any of the four computer-supported tools resulted in a significantly smaller navigation time than manual backtracking using the Table of Contents. Subjects using either of the entry tools had consistent backtracking times across trials regardless of backtrack task length. When provided with a history list, subjects in the entry condition had significantly smaller navigation times than subjects in the component condition. Users did not show any differences between computer tools in rated efficiency, ease of use, or objective preference measures.