The GPII Shopping Aid – Ordering Choices

Denis Anson, MS, OTR

Misericordia University, Dallas, PA
Raising the Floor – US, Washington, DC


The purpose of the Global Public Inclusive Infrastructure (GPII) is to ensure that everyone who faces accessibility barriers due to disability, literacy, digital literacy, or aging, regardless of economic resources, can access and use the Internet and all its information, communities, and services for education, employment, daily living, civic participation, health, and safety (Raising the Floor 2011). The mechanism behind this effort is the creation of a system that allows the individual to define a set of access preferences which live in the cloud, then to call up those preferences from any assistive technology, and have the system automatically configure the device to match the individual’s needs.  When the individual is done using the device, it will automatically revert to its base state.  This saves the individual from having to disclose their disability unless it is desired, and assures that the individual will have access to any information technology device that is connected to GPII, even if they have never encountered it before.

The GPII Shopping Aid Selection List

When an individual comes to the Shopping Aid, they provide, through one of several mechanisms, a list of accommodation needs (Without something like this, I can’t use the device.) and wants (If the product has this feature, I will be more likely to use it.).  This needs and wants list is used to generate a list of products and services that address one or more of the needs of the individual.

 When only a single or a very few choices are available, the order in which choices are presented probably doesn’t matter.  But, when there are several, possibly dozens of choices, order does matter. Each product in the list will address one or more of the user’s needs, but two different products may address one or more of the same needs, but also address different specific needs as well.

Users are more likely to select products near the top of the list simply because they are at the top.  It behooves the developers of the Shopping Aid, therefore, to consider how best to order choices, and to avoid any concern that the order might somehow be coerced to meet the needs of a powerful vendor rather than the assistive technology user.

The process of ordering a selection list from “best” to “worst” requires a definition of “best,” and the development of a means of computing, from available information, what constitutes “best.”  This paper describes the components that the GPII Shopping Aid will consider in determining the “best” product, or set of products for an individual user. 

Calculating the Best Solution

Utility Cost

Clearly, the cost of a product is a major consideration, but it is not simply the retail price of the product or service.  The “Utility Cost” of a product is the total price of the product divided by the number of (unmet) needs of the individual that it addresses.  This cost varies with potential users, because each user might be considering the product for a different set of needs that it addresses.  In one case, an individual might be considering a product costing $300 because it addresses two specific needs, for a utility cost of $150.  Another individual might find that the same product addresses three of her specific needs, so for this person the utility cost of the product is only $100.

For many individuals, the needs profile may require a combination of several products to address all of the indicated needs.  In such a case, the Utility Cost of a product may change through the decision making procedure.  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, resulting in 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 addressed need.  In the initial listing, the utility cost of product 1 is lower than that of product 2 ($100 compared with $110), so will be sorted higher on the listing.  If the user selects product 1 as part of her accommodation profile, the utility cost of the other products on the list must be recalculated.  Product 2 replicates the accommodation for need D, so now addresses just unmet needs C and E, for a utility cost of $165 per unmet need.  After each product selection, the utility cost of available products will change, and the list must be reordered.

Focus Cost

Some accommodations are highly focused, addressing very specific needs.  Others are very broad, attempting to meet a wide range of needs for as diverse a population as possible.

Some types of accommodation can be classed as “appliances.”  That is, once configured, the user does not interact with the accommodation, they act through it.  An example of an accommodation appliance would be setting foreground and background colors to maximize reading for a person with visual limitations.  In most cases, the colors are set once, and persist.  The user does not use one color set in one application, and another set in another application. 

Other accommodations are “tools.”  The individual actively uses the accommodation to enable task performance.  On-screen keyboards and screen readers are examples of AT tools.  The individual interacts with the on-screen keyboard to provide information to a device or application, or issues commands to the screen reader to indicate what text should be read, to spell a word that is difficult to understand, or a host of other tasks.

The demands that these two types of accommodation make on the user are quite different.  Once an appliance is configured, the user ignores it, and benefits from the services provided.  The demands are very small.  On the other hand, a tool may be designed so that it works smoothly, with little effort, or may require substantial effort to perform its tasks.

While the intrinsic nature of an accommodation is fixed, the nature of products is not.  When a product addresses a single need, its interface can be simple, providing just the controls needed for that accommodation.  On the other hand, a product that attempts to address multiple needs must provide the interface for each of the accommodations provided.  If the product accommodations match the needs of the user, this can result in a single, integrated interface that meets the individual’s needs.  The alternative might be to have three different products, each with a different, and possibly overlapping, interface for the user to learn. For an omnibus product (a product that attempts to provide solutions to the widest possible range of needs, so that it can help anyone), the interface may involve controls to many services that the user does not need, resulting in unnecessary complexity.

In main-stream products, there is a tendency for manufacturers to add as many features as possible – to tick all the boxes – in order to make the product more desirable.  In AT accommodations (and in some main-stream products) this can result in a product that is needlessly complex and difficult to use.

The “Focus Cost” of a product is a function of the cost of the product multiplied by the number of features provided that are not needed by the individual.

Desirability Cost

In the selection of accommodations, not all features rise to the level of “need.”  However, the degree of dissonance between the approach to accommodation taken by a product and the desires of the individual must be taken into account.  For example, a word prediction product can provide a prediction list that follows the insertion point, so that it is always at the focus of the user.  Alternatively, the prediction list can be provided in a fixed location, so that the user is not unnecessarily distracted by it while typing.  Different users may prefer different types of prediction list, but be able to benefit from either one.

If the user has generated a list of desirable features, in addition to needs, the Shopping Aid will computer a “desirability cost” for each product that is the cost of the product divided by the desirable features that it provides.  The desirability cost may be somewhat more difficult to make meaningful than the utility cost.  Consider a person who is seeking accommodation for low vision.  This person may have identified three desirable features in her accommodation: the ability to select the voice that is used to read aloud, the ability to have high-contrast display, and having a low-force keyboard. Of these features, two apply to information display (one to vision, one to hearing) but the third applies to the hardware.  A screen magnifier that also can read aloud might gain desirability because of the voice selection and color pallet features it provides, but should not be penalized for not providing low-force keys.  Similarly, a keyboard might have high-contrast, large print keycaps, and might also be low-force, making it desirable, but not provide a choice of voices. 

This reasoning indicates that the desirability index should be additive, not punitive.  Products should benefit from providing desired features, but not be penalized for not providing them.

Available Support and Service

Each individual will have a unique level of support available.  From the path taken through the Shopping Aid to the product listing, we can deduce whether an individual has access to an AT specialist, a clinician who is not an AT expert, or only community/family support. 

Some assistive technologies are easily installed, and have little need for after installation support. A sticky-keys application, for example, may need to be turned on within an operating system, and then will provide its accommodation automatically.  The only likely problem would be for it to be turned off accidentally. Other technologies require more careful installation, substantial training, and may require on-going support.  An AAC product, for example, must be configured to the needs of the individual, the individual may need many hours of training to become a fluent communicator, and ongoing vocabulary maintenance may be required.  Even within a single category, degree of difficulty may vary substantially.  Screen readers like NVDA or JAWS are much more powerful than the ThunderStorm screen reader, but also are much more complex to use.

Each product will have an associated “installation complexity” and “support demand” metric.  These will be used for both exclusion of inappropriate products and sorting products within categories.  If, for example, the Shopping Aid has determined that an end-user does not have professional support available, products that require careful installation and on-going training and support might be excluded for the list.  In these cases where support is available, products with lower support needs will be rated more highly than those with high training and support demands.

Shopping Aid Product Lists

When presenting an individual with a suggestion list based on the identified needs and wants, the Shopping Aid will combine the Utility Cost, the Focus Cost, the Desirability Cost, and the Support Cost to determine the list order of potential solutions. The precise algorithm of sorting has not yet been determined, but our first approximation can be described.

 The strongest factor in sorting will be the Utility Cost.  Products that provide the highest benefit for the lowest cost will be listed first in the Shopping Aid.  The interaction between Utility Cost and Focus Cost has not yet been determined, but we know that a product that has a high focus cost (a large number of unneeded features or relatively high cost for the unneeded features) will rank lower than one with a lower Focus Cost.

Desirability Cost will probably add to Utility Cost (being more desirable plus being more effective for the price).  However, a product that has a high level of desirability cannot rank above one that has a high utility. 

Of these factors, the only one that can trump Utility Cost is Service Cost.  A product with high utility and high desirability, but is complex to install and requires on-going training and support cannot be recommended to a person who has no access to professional or family support, and must install and maintain the product themself. 


The GPII Shopping Aid, when it becomes available, will allow AT professionals, clinicians, and individuals with community or family support, or working alone, to describe their needs, and will suggest only those products that address those needs.  This filtering of the comprehensive Unified Listing to individual needs will minimize the risk of information overload (Eisenstein, 1979; Mumford, 1970; Palme, 1984; Wallace, 2007). 

This paper identifies factors that will be used to provide, at each point in the product selection process, the “best” choices.  The Shopping Aid will not make choices for the individual, but will provide a carefully tailored listing of products among which to choose.  The individual can then explore and make decisions from this list.  But to provide the best possible solutions, the most effective choices should rise to the top of the list.

The factors being considered by the Shopping Aid are sufficiently complex that a clinician, or even an AT expert, will benefit from the ordering, as it is unlikely that most would be able to weight these factors “on-the-fly.” 

The list-ordering feature of the Shopping Aid is one example of how the tool will benefit users with all levels of expertise in assistive technology and disability.


Eisenstein, E. (1979). The printing press as an agent of change: Communications and cultural transformations in early-modern Europe. Cambridge: Cambridge University Press.

Mumford, L. (1970). The Myth of the Machine (Vol. 2). New York: Harcourt, Brace.

Palme, J. (1984). You have 134 unread mail! Do you want to read them now? In H. T. Smith (Ed.), Computer Based Message Services. North Holland: Elsevier.

Wallace, D. F. (2007). The Tsunami of Available Fact. In D. F. Wallace (Ed.), The Best American Essays 2007. New York: Mariner.


This research was funded by the European Union's Seventh Framework Programme (FP7/2007-2013) grant agreement 289016 (Cloud4all) and 610510 (Prosperity4All), by the National Institute on Disability and Rehabilitation Research, US Dept. of Education under Grants H133E080022 (RERC-IT) and H133E130028 (UIITA-RERC) and contract ED-OSE-12-D-0013 (Preferences for Global Access), by the Flora Hewlett Foundation, the Ontario Ministry of Research and Innovation, and the Canadian Foundation for Innovation, by Adobe Foundation, and the Consumer Electronics Association Foundation, with support from IBM, Microsoft, and the participating partners.   The opinions and results herein are those of the authors and not necessarily those of the funding agencies.

Audio Version PDF Version