Brumby et al. (2014) looked at how knowing what you are searching for or not on a menu affects search behaviour:
- p's instructed to search a menu given the word (known item search) or just given a semantic cue (i.e. the category the word belonged to) for the word (i.e. semantic search)
- using eye tracking, found that p's were faster and more accurate at searching in the known-search condition --> shown to be because p's skip over items in known search but during semantic search are more likely to assess in turn
- p's did not have longer gaze for items in the semantic search task unless the items were arranged close together (close proximity allows to assess multiple items with a single fixation?) --> contrast previous reseearch that assessments of items should be longer during semantic than known-item since it takes time to retrieve info from memory (Young et al., 1998).
- people adopt different eye gaze strategies depending on the type of search activity they are engaged in
- future research should model the data to explore them in more detail
Label relevance for assessment and selection:
- When searching a novel Web page, estimates of label relevance play a substantial role in determining link selection (Chi et al., 2003) --> during a search the user must decide which items to assess - estimates of relevance must be embedded within a strategy for controlling search.
- Interactive search = searching a novel Web page for information that is relevant to the achievement of a particular search goal (Payne, 2000)
- Rieman (1994) p's showed a label-following strategy while learning a novel-graphing package --> tended to select items from the application menu with labels that had a high overlap with the current goal
-p's often would not assess all items in the available choice set and would repeatedly reassess (and invest more time in each assessment) a small subset of those items that were initially assessed
- later characterized this search strategy as an iterative deepening of attention -->
ps should spend more time on each successive revisit to an item due to increasingly high-quality and costly assessment methods
Several account of how people control search:
- In some accounts, assumed that people consider all items on a page prior to making a selection (advantage that it guarantees that the label with the highest likelihood will be selected).
- in others, assumed that people make a selection immediately following an assessment of a highly relevant item. (more like Simon’s, 1955, satisficing heuristic advantage that a good-enough label may be found in less time).
Brumby & Howes (2008) wanted to discriminate between assess all and satisfiicing accounts of interactive search.
- 2 expts systematically manipulated the relevance of the distractor items and the location of the target item on the page --> eye-tracking method used to determine the number of items in the set of options that p's tend to look at prior to selecting an item.
- 1) high value items were more likely selected when it was first encountered if the relevance of competing distractors was relatively low
-ps were actually more likely to select an item immediately after visiting it for the first time when the distractors were less relevant (consistent with accounts that assume that people adjust an estimate of the likelihood that a particular item will lead to the goal given the context provided by the likelihood that other items in the set will lead to the goal) --> does not support Rieman et al.’s (1996)
= people are more strategic and sensitive to context than previous models of interactive search suggest
- 2) More items were assessed prior to selection when the distractors were of greater semantic relevance to the goal
- sensitive to the context provided not only by low relevance distractors but also by the context provided by highly relevant but as yet unselected items. Although people may not commit to an immediate selection of a highly relevant item, the assessment of such an item does have consequences for subsequent assessments. (none of the current models of interactive search provide a direct explanation for this pattern of item skipping behaviour).
- because it was found that the decision to select the target was shown to be dependent on its position within the set of options, these findings do not support the hypothesis that people assess until the most recently assessed item exceeds a threshold.
--> inconsistent with both satisfising and assess all accounts of interactive search
Length - whilst longer menus lead to longer search times (i.e. because takes longer to make a decision), as one evolves from a novice to an expert a strategy shift takes place, resulting in the loss of dependency on linear visual search
Bailly et al. (2014) extended Nilsen's work --> looked at effect of menu organisation, absence and presence of target item, target position and practice, as well as menu length on item selection.
- Similar procedure but menus of 8, 12 or 14 items, split into either unordered, alphabetical or semantic. Utilised four blocks per menu = examine practice effects.
- Replicated findings --> more items on menu = slower ps were at selecting an item
- Results for first block for each menu replicated - response time increased with serial position. BUT effect only apparent for first block of the item, then for subsequent blocks the effect diminished (i.e. RT no longer increased with serial position) --> PRACTICE EFFECT (learnt the position of the item)
- Learning also seen on trials where the menu didn't contain the target (i.e. p's press space bar to indicate its absence), where initial slowness diminished with practice
- Unlike Nilsen saw 'last item effects' --> last items on the menu were selected faster than previous items.
- Explained using their model which hypothesises that gaze distribution is normally distributed, centred on the target = when last item is the target, predicted gaze distribution is cut off for items after the menu, reducing the amount of gaze for the target
- Also found p's able to select items faster when menus organised alphabetically or semantically compared to unordered organisation (also diminished with practice) --> link to menu organisation
Cockburn (2007) effect of learning on the effect of menu length
- provides a predictive model of menu selection time, utilising Fitts' Law (the time it takes to reach a target varies as a function of the ratio between the distance to the target and the width of the target) and Hick-Hyman law (decision time varies as a function of amount of info conveyed by a stimulus)
- Based model on observation of Sears and Shneiderman (2004) --> as menu length decreases, selection times decreased linearly for infrequently selected items but decreased logarithmically for frequently selected items
- = upon learning, p's no longer rely on linear visual search, they started using Hick-Hyman decision times once they had become familiar with item locations
- their model incorporates the transition from linear visual search to logarithmic decision time as expertise is gained. Testing this model confirmed this strategy shift: for novices search time increased linearly as item position increased, however for experts this relationship was logarithmic.
- model should apply equally well to any interaction involving a choice decision followed by a pointing task e.g. selecting desktop items
Classic study by Nilsen (1991) --> p's search for a specific digit in a menu of randomly-ordered digits, using menu lengths of 3, 6 and 9
- p's response times approx. linear function of serial position
- Each successive position required approx. 100ms longer to select than the last.
- RT slightly longer for the first item in the menu (expect it not to be first??) and the longer list = longer to search
Might depend on situation: e.g. Kuhn in car device interaction
- Interacying with an in car device = reduction in driving performance but accidents are not more common because of this = do drivers adapt their device use to ensure safe driving?
- p's controlled simulation of car in city or on highway and eye tracking data taken whilst interact with MP3 player
- City driving --> slower glances at in-car device - greater demand than highway?
- Highway driving --> made longer glances
- no difference in number of glances made, just duration
- Adaption in city maintained good driving performance = adapt interaction with a menu depending on the situation
Organisation- unordered menu organisation is detrimental to search time. Stable menu organisation is more advantageous than a changing menu organisation, as it facilitates learning and consequently faster selection.
Menus organised alphabetically/semnatically = faster location and selection of an item compared to disorganised menus:
- So much evidence for this e.g. Card, 1982; Halverson & Hornof, 2008; McDonald et al., 1983; Mehlenbacher et al., 1989).
- General explanation = item positions are stable and so easily learned
- But, distinguishing between the 2 types of organisation is difficult --> found that for targets with unknown labels, menus that are organised semantically = faster search times than menus with alphabetical organisations (McDonald et al. 1983)
- Contrary- Bailly et al. (2014) investigating item selection with targets with known labels has produced mixed results
- overriding result is that alphabetic organisation produces faster search times (e.g. Card, 1982; Mehlenbacher et al., 1989)
Chen et al. (2015) proposed a model in order to explain the conflicting data
- Suggests that one’s behaviour changes depending on the situation --> called 'strategic adaptation'
- Doesn't predict faster performance for either alphabetic or semantic organisations unless training occurs --> can create an advantage for one type of organisation over the other, whilst causing worse performance for the type of organisation on which the participant was not trained
= one adapts one’s strategy based on the menu type.
- Predicts that gaze data will demonstrate that semantic and alphabetic organisations are superior to unorganised menus.
- When applying the model to a real world distribution of menu items and comparing it to human data from a previously reported experiment, nearly all of the predictions of the model were supported, however the model predicted a more beneficial effect of menus organised alphabetically than was observed.
Organising by frequency of use of each item:
- research suggesting that objects at the top of the list are quickest to find and select (Bailly et al. 2014) suggests that this method of organisation would be beneficial in reducing search time for freq. used items
- Sears and Sneiderman (1994) found support for frequency organisation over alphabetic organisation
- Cockburn et al. (2007) found support for faster menu search with menus organised by frequency compared to other types of split menu e.g. recency of use.
- BUT, freq. of use of items changes constantly = prevents users from learning where certain items are in the menu which could be detrimental to the process of searching (shown by Bailly et al) --> practice effects overruled other organisation effects = plausible that stable menu organisations (alphabetic or smenaitc) which promote learning would --> faster menu search than unstable menu organisations such as split menus
- Almost every computer tech involves a menu of some sort = HCI researches have conducted a large amount of work on how users interact with menus
- increasingly important to design interfaces that entail a positive user experience and menus are essential components facilitating the user to navigate through different functions to achieve their goal
- Brumby and Zhuang (2015) regardless of the application domain, menus should allow the user to locate and select the desired item with ease and speed = the exploration of how one can design a menu in order to minimise errors and maximise speed during item selection is very important
- key technique used is eye tracking to see how attention is given to a menu and there's been a lot of consideration to menu structure
The visual system:
- Keith Rayner, cognitive psychologist who pioneered modern eye tracking methodology in reading and visual perception
- When we read we fixate (momentary pause) and do saccades (ballistic movement to new location) --> often skip over words (high freq. ones usually) and our eyes land (fixate) on parts that contain the most information
- Rayner (1974) claims to be able to recognise a word it must be within 3 to 4 character spaces of fixation
Disadvantages to consider:
- Eye tracking is noisy, and distraction, boredom and level of motivation may affect how a person moves their eyes when interacting with a menu
- Most studies look at linear menus --> not representative of all menu interfaces we interact with on day-to-day e.g. phone or laptop home page menu
Pan et al. (2007) too much trust placed on google
- Eye tracking method with college students. Found they have substantial trust in Google's ability to rank results by their true relevance to the query
- When p's selected a link to follow they were strongly biased towards links higher in position even if the abstracts themselves were less relevant --> large implication for influence of Google on society and user traffic on the Web
- Early research suggested that Web users are functionally blind to rectangular graphics that they perceive to be advertisements
- but more recent studies indicate that people do notice ads, dislike them, and that site credibility suffers - Fogg et al. (2001)
- Pagendarm and Schaumburg (2001) propose that “banner blindness” may occur especially in a goal-directed task
Burke et al. (2005): - P's instructed to search menu for either known headlines or a semantic cue of the headline
- adverts affect search times in both conditions
- eye tracking showed p's hardly ever looked directly at the adverts (i.e. showed banner blindness) = were distracted by adverts through peripheral vision
- on subsequent memory test saw low banner recall and adding graphics didn't make the averts more attended to
- should post ads closely related to page content if want to attract attention --> connects advertising to viewers goals
- ads should not be placed in predictable locations - p's have learned to avoid these areas
- to lessen the spread of banner blindness to critical page elements, usability guidebooks (i.e., Nielsen and Tahir, 2002) advise against placing site navigation above banner ads
- on menus in important situations (e.g. a hospital), no info should be placed on the side of the menu (in banners) to avoid distraction and so reduce search times