Shopping Aid

From wiki.gpii
Revision as of 23:34, 5 June 2014 by Stschmitt (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

User Needs Analysis

Use Cases

There are a number of situations where the GPII Shopping Aid can be used to assist in matching the needs and preferences of the individual to the available assistive technologies.

The End User with an Expert Clinician

In the presumed ideal case, the individual needing assistance to access information technologies has the assistance of an expert clinician in making choices. The expert clinician, in this scenario, is able to evaluate the skills and abilities of the individual, the demands of the task for which assistive technology is being considered, and the constraints of the environment or environments in which the task is to be performed. Combining this information with expertise on the impact of various types of intervention, the expert clinician formulates an image of the features of an ideal solution. This solution may or may not exist in a single product, but is a conceived as a set of functionalities. The expert-clinician oriented interface of the Shopping Aid will be designed with a features orientation. The clinician can specify that the solution requires reading text aloud, with simultaneous highlighting of each word as it is spoken. In another case, the clinician may specify a smaller keyboard, with full travel rather than membrane keys. In a third case, the clinician may be looking for a system to increase the size of text on the screen, to provide a larger keyboard with tactile cues, and abbreviation expansion capability. The ultimate solution may be included as system settings adjustments, or the addition of one or more add-on assistive technologies. The second step of the process is locating the technological solutions that will provide the functionalities required by the end-user. Even an expert clinician cannot remain current in all the available assistive technology products, the built-in solutions, and custom solutions in the field. New products are constantly introduced, existing products change their features or are withdrawn from the market. The GPII Shopping Aid assists the expert clinician by presenting constantly updated information on available products and product information. This allows the expert to consider new or revised products that might meet the needs of an individual.


The End User with a Clinician who is not an AT Expert

While experienced or well-educated clinicians are able to evaluate the skills, abilities and limitations of the client, perform an activity analysis of a task and demands analysis of the restrictions of the environment, they may not have extensive experience with assistive technology. The “clinician interface” of the Shopping Aid is designed specifically for such users. The clinician interface, rather than focusing on the features of assistive technologies (as does the developer interface) accepts, as input, the functional needs of the client. The clinician will know, based on their evaluation of the client and the task, what performance barriers exist for this client. For example, the clinician may have observed that the client needs to be able to type the full alphabet, but becomes confused when more than 12 keys are presented at once. The clinician may not be aware of any assistive technology for such a case. If the client has good motor skills as well, the Shopping Aid might recommend a chording keyboard. If the client’s motor skills are lower, a switch encoding system such as Morse Code might be recommended. It is not necessary that the clinician know that such accommodations exist to receive them as recommendations, so long as the functional needs and limitations of the client can be described.


The Expert Clinician Without Direct Contact with the End User

Expert clinicians are often called upon to consult in assistive technology provision. In such a case, the end-user may have been evaluated by clinician who is not an AT expert to determine the user’s skills and abilities, and the requirements of the job. The expert clinician may never do a personal evaluation of the client. In other cases, using tele rehab, the expert clinician may use a videophone connection to observe the evaluation and to suggest additional assessments, but does not have “hands-on” contact with the end user. In such a case, the structure of needs entry into the Shopping Aid can help structure the provision of information, and improve the outcomes by assuring that all relevant issues are considered. As the clinician provides functional needs information to the Shopping Aid, the pattern of data entry cues questions between the expert and non-AT expert clinician, facilitating full consideration of options. Such cuing is important because, in many cases of hands-on evaluation the expert is gathering subtle signs that they may not consciously be aware of. The structure of data entry of the shopping aid will assist in bringing these subconscious considerations to the fore.

The End User with Family Assistance

Many people in need of assistive technology may not have access to clinical assistance in their search. Whether because of financial limitations, where the individual lives, or simply because the individual is not currently “sick,” professional assistance may not be available, or, in many cases, even appropriate. If an individual is having a hard time using an electronic device. they may not entirely understand why they are having difficulty. In most cases, they will not be able to discover or describe the aspects of using a device that are causing difficulty. At most, they know that the device is hard to use. Many will blame themselves for their difficulty, rather than seek an alternative device or interface changes. In some cases, the individual will be able to recruit the assistance of a family member or friend in seeking a solution. While this family member may know little more about the limitations that are making technology hard to access, they may be more aware of how access is supposed to work. Consider, for example, a person who is having a difficult time using an electronic timer because the alarm frequency matches the high level hearing loss of the elder. While it is possible to use the timer by watching the digits count down, this is of little utility. When seeking help from a family member who can hear the alarm, the reason for the problem becomes immediately apparent. Once the problem is identified as not hearing the audible alarm (which the person with the disability wasn’t aware of), the family member and end user can seek a timer with an alarm that can be heard. When using a computer, the end-user may be frustrated because the computer suddenly stops working, for no apparent reason. A family member with better eyesight might be able to see that the program has asked for input, but that the request is hard to see because it is too small, low contrast, or on the far left of the screen. Again, knowing what should happen helps identify the problems that are occurring, making the solutions easier to find. An on-line aid with the purpose of assisting access for an individual to get on-line faces a profound chicken-and-egg problem. Until the individual is able to get on-line, the assistance provided on-line is of little use. But recruiting a more-able family member to assist provides the support necessary to begin the process.

If the user and family member are unable to provide sufficient detail, the Shopping Aid may suggest that the end user seek the assistance of a professional to more clearly identify the needed assistance.

The End User Alone

The most difficult to manage case for the GPII Shopping Aid is the end-user accessing the system alone. However, this case is also that of a large, but specialized group of potential users.

This use-case is difficult for two reasons. First, the user is attempting to find the AT or services that will allow use of an ICT device, using an ICT device. In some cases, the user may be seeking information about systems that are not currently fitted with AT, from one that is, so this is not a complete impasse, but in many cases it will be difficult. In order to accommodate such users, the interface for the Shopping Aid will have to include a screening for access methods, and adopt the methods of input and output selected by the user (so far as those methods are available without additional software or hardware on the host device) as soon as they are selected. The second difficulty is that, while people without clinical training may be aware of having difficulty with controlling or perceiving an information or communication device, they may not be able to adequately describe the nature of the problem they are having. Without prior experience with how the interaction “should” be accomplished, they cannot identify the areas where they have difficulty. In addition, they may not have the vocabulary to describe what they are experiencing. Most non-medical people describe any vision issues, for example, as either “needing new glasses” or “double-vision.” They simply don’t have the vocabulary and concepts to describe more complex difficulties with visual perception.

To accommodate the needs of end users working without assistance, the end user might first use the Discovery Aid. This part of GPII is designed to help the end-user discover the areas where they are having difficulty, and automatically pass those needs to the Shopping Aid. The Shopping Aid will then offer s concise list of products and services that might assist the end user in performing their tasks. The Shopping Aid may also indicate that some individuals with similar needs also find additional solutions useful, and offer to recommend those additional solutions.

Alternatively, the Shopping Aid will include a mini-assessment that the more advanced user might find helpful to identify needs. In the majority of cases, individuals require only basic assistive technologies, and these can be identified without complex assessment.

If the user is unable to provide sufficient detail, the Shopping Aid may suggest that the end user seek the assistance of a professional to more clearly identify the needed assistance.

Wireframing

Shopping Aid “Unsearch” [Reverse search ??]

The purpose of the Shopping Aid is to assist end-users and clinicians in making superior choices in the selection of assistive technologies to facilitate access to information and communication technologies (ICT). The assumption of this process is that, at the beginning, the user does not know what the best assistive technology will be.

With this assumption, it can be seen that a conventional “search” feature, where an individual seeks information about a specific product is of little utility. If the individual knows the product that they want, they could more efficiently obtain information directly from the manufacturer.

Clinicians working in assistive technology may fall victim to Maslow’s Hammer. In 1966, Maslow wrote, “I suppose it is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail.” (The Psychology of Science, 1966) For a clinician who only recommends assistive technologies occasionally, or an end-user with limited exposure to options, there is a tendency to use the known assistive technologies even when better, but unknown options exist. An assistive technology product may include a secondary feature (text to speech in a word processor), and that secondary feature may provide the accommodation that the user needs. But an application that provides that feature as its principal function may well provide a more versatile version with a simpler control set. There is an application of search technology, however, that may have utility for the clinician or end user. That is the ability to search for things “like” a specific product, provided the user can identify the ways in which “likeness” is to be defined. For purposes of discussion, we will call this operation “unsearching,” because rather than moving from the general to the specific in a conventional search, we will be moving from the specific to the general. We will need a better name for this feature, on that makes its function clear, before it is released.

The Unsearch Process

In order to find products “like” a know product, the user must follow a two step process. The first step is to identify the product by typing its name into a search box. If there is only one product with a name like that provided, the user will move directly to the second step. It is also possible that, like a conventional search, there may be more than one product that matches what the user has typed. In this case, a list of likely candidates will be provided for the user to choose among. This list might include the formal product name, manufacturer, and link to the product page, so that the user can find the specific product being considered. Once a specific product is identified, the user moves to the second phase of “unsearch,” identifying the characteristics of the product that are important for the current application.

Each product in the Unified Listing will have fields identifying the functionality of the product. These are necessary to allow matching of identified need to products. In the case of “unsearch,” once the user has specified a product, the screen will show a list of functions of that product. The user will be asked to identify which features of the product were being considered as making it a potential solution for the current case. In many cases, especially for “omnibus” products that provide a wide range of features, only a few of the available features will be considered important.

When the user submits this list of functionalities, the Shopping Aid will respond with a list of products that provide those functionalities. Each product will be followed by a grid, with the specific functions provided by that product, and a “goodness of fit” number. This number will be, in part, derived from the number of functions of the product being used divided by the total number of functions provided. An additional factor will be the number of indicated needs that a specific product addresses. A single product that provides all of the needed functions, and nothing else would be preferred to a product that meets all of the desired functions. It will also have a cost factor, since it is likely that two relatively inexpensive programs that work well together provide a better solution than a single expensive program with lots of unused features.

The list of products should be sorted by the goodness of fit metric. This way, the “better” solutions will migrate to the top of the list. I’d like to figure out a way to automatically recommend combinations of products that work well together and meet the overall needs of the user. This way, it would be possible to indicate that one product that has a goodness of fit of .4, combined with another that has a goodness of fit of .5 together provide a better solution than a single product with a goodness of fit of .8.

Product Selection

Because many of the details that drive technology satisfaction are idiosyncratic, the Shopping Aid cannot, and should not, provide a single recommendation in most cases. Thinking style, aesthetics, and command structure may be the final determinants of the “ideal” solution for an individual, and we cannot know exactly how an individual will respond to a specific product. Ideally, the Shopping Aid will provide an option for “try before you buy,” so that users can explore the options that are available.


Thoughts on entry to the shopping aid

For an end-user coming to the Shopping Aid unassisted, there are four identified conditions.

  1. Transfering from the Discovery Aid
    1. This user has worked through the process of the Discovery, and the aid has identified the primary user needs.
    2. Because we know the user’s identified needs, we are able to provide an UI that works for the user.
      1. For example, if the user is blind, we can begin with a self-voicing interface
      2. For a person who has only a single reliable motor action, we can begin with an autoscan interface.
    3. Because the transfer process includes the list of identified needs, we can move directly to finding candidate products for the user.
    4. The list of candidate products will be arranged by the “Goodness of Fit” calculation, with the better fitting solutions at the top of the list.
  2. Current GPII user, returning to the Shopping Aid to update their product choices
    1. This user has a GPII preference file. The preference file includes the identified needs, as well as current products and their settings
    2. There are two possible interface conditions that might cause the user to return to the Shopping Aid
      1. Because of changes in the user’s function, the current devices/settings are usable, but are not satisfactory
        1. It is possible that the user has been alerted to new updates or products, and wants to know more
        2. It is possible that, over time, the user’s abilities have changed, and stronger magnification, different response speeds, or other changes are needed.
      2. Because of changes in the user’s function (abrupt, or cumulative), the user is not able to use the current AT solutions.
        1. In this case, the path will be similar to that of the person who we don’t know anything about, except that we have a “last known” workable solution. If the person has low-vision, for example, it is unlikely that their sight has returned. Much more likely their condition has worsened, but we can start with what we know rather than from scratch.
  3. The individual is a current user of assistive technology, but new to the shopping aid
    1. There are four scenarios that could apply 
      1. The user is happy with current AT, but would like the ability for the AT configuration to follow them, and appear on new systems
        1. The user is satisfied with his/her existing AT products, but needs to adjust the settings to provide a better user interface
        2. The user needs to update to a current version of a product that is already in use
        3. The user is dissatisfied with the current assistive technology, and wants to find a new product that will be more satisfactory
          1. In this case, “Unsearch” would provide a list of similar products that potentially meet the user’s needs
    2. It would be useful if the shopping aid or GPII interface could identify the currently running assistive technology to determine what products are in use, and what their settings are. 
  4. A new user with unknon needs, but who does not want to use the Discovery Aid
    1. This user may prefer, and be able to, just describe the needs
    2. The shopping aid might ask questions about possible difficulty in seeing, hearing, controlling, or understanding aspects of the technology
      1. The user might be able to say, “Yes, I can do that,” “No, I can’t do that,” or “I don’t know.”
      2. We might do a trichotomous question process:
        1. Yes I can do that
        2. I need help with that
        3. I can’t do that.
    3. The first challenge is to find an avenue of interaction with the user: The user is able to provide information to the shopping aid, and the information from the aid is perceptible to the user.
      1. This interface has to be usable, but need not be optimal. It can be used to find a more optimal solution
    4. Once a method of interaction is established, the user is given the opportunity to identify or discover access issues
    5. When the needs are identified, the user can move to the product selection process.

Product Ordering in the Shopping Aid

When an individual comes to the product listing of the Shopping Aid, they will have expressed or submitted a set of needs and wants. These needs and wants, translated into functions, will be used to generate a product list from the Unified Listing. The selection criteria will be that the product functions match at least one of the needed functions, and may match one or more of the wanted functions or characteristics.

For each product in the list, a “utility cost” will be computed. The utility cost is the total price of the accommodation divided by the number of (unmet) needed features that it provides. This formula balances the cost of products with their versatility. A low cost product that provides just one functionality may not be a better selection than one that costs 75% more, but meets two needs. A moderately priced product that meets several of the individual’s needs is a better choice than a similarly priced product which has even more features, but meets fewer of the individual’s needs. Once the user has made a product selection, the utility costs must be recomputed and the list reordered. The reason for this is that utility cost must be based on the cost for needed features. Consider the case of a user with five identified needs, A through E. Product 1 addresses needs A, B, and D, for a cost of $300, or a utility cost of $100 per need. Product 2 also addresses three needs, C, D, and E, for a cost of $330, or $110 per need. If the user selects product 1, the utility cost of other products must be recalculated. Product 2 replicates Need D, which has been met by Product 1. Therefore, Product 2 now provides for only two unmet needs, so the actual utility cost, once Product A has been selected, is $165 per unmet need.

In addition to utility cost, there should be a focus metric. A single function product will have a very focused user interface, and, if well designed, be easy to understand and use. A product that provides a wide range of services will, by necessity, have a more complex interface, to allow control of each feature. A product that provides only features that are of interest to the user would generally be a better fit than one that provides many features, of which only a few are relevant. The focus metric, expressed as a ratio of the number of needed features plus the number of wanted features (see below) provided over the total number of features provided would be a measure ranging from 1 to a small fraction, if say, only 1 of 30 features were actually needed. There would likely be a scaling factor to properly compute the desirability of the product. This scaling factor assures that products are not punished excessively for having a few unneeded features.

A third metric, “desirability cost” will also be computed when the user has specified wants as well as needs. Like utility cost, the desirability cost will be product cost divided by the number of desirable features that are provided. The desirability index is somewhat more difficult to make meaningful. Consider a user with low vision who had identified four desirable features. These include selecting the voice to be used to read aloud, selecting from the full color pallet rather than just from a small set of high-contrast sets, and having a keyboard that is low-force. Of these three desired features, two apply to information display (one to vision, one to hearing), and the third applies to the hardware being used (the keyboard). A screen magnifier may allow color selection and voice selection, but would not provide a specific keyboard feel. A high-contrast keyboard would not provide the voice or display options. Would a low-force, velvety touch keyboard be less desirable because it doesn’t allow color selection? On the other hand, all keyboards being considered would have the same potential to provide wants, so this may not be an issue.


The default order of product listing will be utility cost, from low to high. The user will also have the option of sorting the list by functionality, so that all screen magnification products, for example, rise to the top. In this case, the secondary sort (within screen magnifiers) would be by utility cost. The tertiary sort order in all cases will be by desirability. This means that, when suggestions are sorted by feature, the screen readers would be sorted so that those that have the lowest utility cost will rise to the top, but when the utility cost is the same (we may make this a fuzzy sort), the more desirable products are listed first. In the fuzzy sort, the difference between the most expensive and least expensive products in a group might be divided into subgroups (deciles or quartiles are likely candidates), and the items in that subgroup sorted by desirability cost.