Publication: Cartography and Geographic Information Science
Publication Date: 01-JUL-07
Delivery: Immediate Online Access
Author: Hope, Sue ; Hunter, Gary J.
Article Excerpt
Introduction
With the growing use of Geographic Information Systems (GIS), the applications of spatial information have become more sophisticated and diverse. Decisions are increasingly being made on the basis of information that is derived from GIS, and the users of these systems may have little or no knowledge of the processes underlying the output products. Although the systems may be computationally precise, the outputs of a GIS can only be as accurate as the data inputs. In recent years, it has been recognized that both the quality of the input data and the resultant effect on the information outputs need to be fully understood if the decisions being made are to be considered robust and informed. However, to date there has been relatively little research conducted into how the inclusion of uncertainty information and the mode of its representation may influence decision-making.
We now accept that errors and uncertainties are inherent in spatial data, and the need for metadata concerning the lineage and accuracy of a dataset is widely accepted. Data transfer standards that include provision for metadata description, such as the U.S. Spatial Data Transfer Standard, have been established in many countries and more recently the International Organization for Standardization has finalized several standards relating to spatial data quality. However, while most spatial data producers now provide metadata with their datasets, researchers such as Hunter (2001) have cited several examples of poor metadata reportage along each of the identified quality elements. He has also emphasized the need for greater detail in metadata reporting (Hunter and Goodchild 1997) and the provision for data quality information at local feature class and feature levels, as well as at the global dataset level (Qiu and Hunter 2002).
Furthermore, while the provision of metadata embedded within spatial datasets is becoming more widespread, their ability to communicate uncertainty is still considered quite limited. Frank (1998) contends that metadata are written from the perspective of the data provider and, as a result, do not necessarily meet user requirements. Indeed, Keuper (2004) found that few consumers of spatial information, including those with substantial experience, give much consideration to the information provided in metadata reports. Thus there is growing recognition, at least within the academic community, that: (1) the functionality of GIS needs to be enhanced to include ways of representing uncertainty; and (2) such representations need to communicate the uncertainty in a manner that is unambiguous, fully informative, and better able to facilitate decision-making.
For that functionality to be most effective we need to know how users of spatial information react to the presence of additional uncertainty information when they are required to make decisions. This calls for cognitive testing in a decision-making environment, and we have sought evidence and indicators from the psychological literature to refine our observations. In this paper we confine our testing to the effects of introducing thematic uncertainty information to spatial decision-making, however, a companion paper has been written that addresses the issue with respect to positional uncertainty.
We begin with a background to our study in terms of the reported research in the spatial information and psychological literature, followed by sections that describe the case study and our method, respectively. We then explain the participating groups used and the way the experiment was conducted, followed by sections presenting the experimental results and discussion.
Background
In the early 1990s, the National Center for Geographic Information and Analysis (NCGIA) established a research initiative focusing on the issue of visualization of spatial data quality (Beard et al. 1991). Research challenges were identified in the areas of communicating data quality to users, error propagation through GIS operations, and the application of data quality information to decision-making--in addition to metadata reporting. Limited progress has been made in each of these areas.
A considerable number of papers have been published on the issue of quantifying spatial data uncertainty, and a substantial amount of research has led to innovative methods of visualizing this uncertainty within GIS. In particular, methods that utilize the advantages of computer systems over paper maps have been developed, such as dynamic representations (Ehlschlaeger et al. 1997), multiple simulations (Crosetto and Tarantola 2001), and, more recently, interactive displays enabling users to query datasets on the basis of quality (Devillers et al. 2005; Pebesma et al. 2006). However, there has been far less research conducted into the effectiveness of these methods as communicators of spatial uncertainty and, specifically, how these uncertainty representations may affect the decisions being made. Indeed, as MacEachren et al. (2005, p. 132) aptly note, "... the research seems to take for granted that visual depictions of uncertainty are useful for decision making."
The relatively few studies that have considered the effectiveness of thematic (or attribute) uncertainty representations have generally been confined to examination of the extra cognitive demands placed upon decision-makers. They have been concerned with the question of whether or not users can cope with the additional information being provided in a display that incorporates uncertainty information together with thematic information (for example, Evans 1997; MacEachren et al. 1998; Leitner and Buttenfield 2000; Aerts et al. 2003). Effective representations have been concluded as being those that do not detract from the accuracy or speed of decisions, and which users rate as easiest to understand. While these findings are important, we contend that the studies have required little application of uncertainty information to the decision-making process. That is, they have tended to test whether users can comprehend such representations, with less consideration given to how the inclusion of uncertainty information might impact on the decisions being made.
An exception was the study by Leitner and Buttenfield (2000), in which participants were required to make decisions concerning the best and worst sites for a new park or airport. However, the correct response was always for the high certainty region. As such, their finding that response times decreased for some of the decisions where uncertainty information was provided may also have been due to participants being able to restrict their search to regions that were labeled as being of high certainty. The case study presented here aims to extend the work of Leitner and Buttenfield and, in a similar decision context, requires participants to not only apply the uncertainty information in choosing between two regions but also to rank multiple regions with different levels of uncertainty.
At this point, a lateral step into the psychological literature provides potential insight into how decision-making may be affected by the introduction of thematic uncertainty information. In attempting to describe human decision-making behavior, models have been developed that are based on the strategy of maximizing expected utility. Early studies considered the question of how participants made a choice between alternatives under conditions of risk, defined as referring to situations where the consequences of a decision depend upon future outcomes having known probabilities of occurrence. The probability of an event's occurrence is multiplied by the value that it offers to provide a measure of its expected value. The option with the greatest expected value would then be accepted to be the rational choice. However, Kahneman and Tversky (1979) discovered that actual decision-making behavior sometimes deviates from this simple case, prompting them to introduce a subjective measure of value, known as its utility.
For example, most people would prefer to accept the sum of $1,000 with certainty, rather than have an 80 percent chance of winning $1,500, despite the greater expected value ($1,200) of the latter. Also, most people display loss aversion, being reluctant to accept a wager where they have an equal chance of winning or losing a set amount. These findings have led to the S-shaped utility function (Figure 1), where utility represents the subjective value of an event. Value displays a diminishing marginal utility, resulting in the curve, and loss aversion causes this curve to be steeper for losses than it is for gains.
[FIGURE 1 OMITTED]
In addition, Tversky and Kahneman (1982) demonstrated that humans are prone to decision-making biases, particularly in subjectively judging probabilities. For example, the probability of occurrence of a recent event or a salient, newsworthy event is often over-estimated. These findings led them to develop one of the most widely accepted models of decision-making behavior: Cumulative Prospect Theory (Tversky and Kahneman 1992). This model introduces a subjective measure of probability in addition to that of value; however, as this model has been refined to accommodate some of the idiosyncrasies of actual behavior, it has evolved into a complicated mathematical function. This function models decision-making behavior reasonably well, but it would appear to be highly unlikely that people are actually utilizing such a strategy in practice.
Instead, the use of simpler heuristics has been proposed to describe actual decision behavior. These are "rules of thumb," such as selecting the first option that meets all requirements. They do not place such high information-processing demands on decision-makers whilst still predicting behavior with a similar level of accuracy. They are also more similar to subjective self-reports of conscious decision-making strategies. Some of the preference reversals that are inconsistent with the strategy of maximizing expected utility could be systematically predicted in terms of the use of heuristics. This gives strong support for the theory...
NOTE: All illustrations and photos have been removed from this article.
More articles from
Cartography and Geographic Information Science
Interpretation and generalization of 3D landscapes from LiDAR data, 01-JUL-07
Relations among map objects in cartographic generalization, 01-JUL-07
Testing the effects of thematic uncertainty on spatial decision-making, 01-JUL-07
Looking for additional articles?
Click here to search our database of over 3 million articles.
Looking for more in-depth information on this industry?
Click here to search our complete database of Industry & Market reports by text, subject, publication name or publication date.
About Goliath
Whether you're looking for sales prospects, competitive information, company analysis or best practices in managing your organization, Goliath can help you meet your business needs.
Our extensive business information databases empower business professionals with both the breadth and depth of credible, authoritative information they need to support their business goals. Whether it be strategic planning, sales prospecting, company research or defining management best practices - Goliath is your leading source for accurate information. |
