Preference Relaxation in Product Search

This paper describes salient issues in cloud computing capable of facilitating innovation in organisations. Focusing on the potential value to organisations Clohessy, Trevor ; Acton, Thomas This paper describes research-in-progress that explores the applicability and implications of cloud computing in the creation of business value through open innovation. Both the cloud computing and open innovation paradigms The importance of effective decision making in organisations has been well documented. Groups are often formed in order to collaborate skills and information and assist with decision-making.

Despite the many benefits associated Most Popular Items Statistics by Country. In this short paper we present our approach using numerical attributes. However, our method can be also applied to other e. Typically, preferences on numerical attributes are expressed using value ranges. Figure 1 Example of edge sets on price preference with items selected for inclusion in the result set squares. The inclusion of all items satisfying the relaxed criteria would significantly increase the number of items presented to the user, resulting in higher information overload.

To prevent these negative effects we incorporate a selection mechanism into our relaxation method that includes only some of those cases.

• You are here.
• Rimo quando calha (Portuguese Edition).
• .
• Files in this item.

To explain this strategy we introduce the concept of an edge set. We conceptualize an edge set as a set of alternatives that fall into a value range constructed based on the initial consumer preference with relation to a given attribute. For every preference value range two edge sets ES can be constructed lower and upper , respectively: More specifically, our approach involves three steps. First, we create edge sets for every interval boundary e. Secondly, for every edge set we identify the set of all non-dominated items also referred to as the skyline Borzsonyi et al. Finally, if any item from the skyline does not satisfy the initial not-relaxed preference it is included in the results set.

To prevent an increase in cognitive load, the total size of the set is kept constant, thus the items with lowest utility according to current preference model are substituted with items from the skyline. Hypotheses We expect that the method we propose will impact various aspects of decision making, in particular decision quality. We propose to measure decision quality by analyzing the consideration sets formed in the decision making process; such sets may comprise dominated or non-dominated superior alternatives.

Sets with largely superior alternatives are better. Thus, we propose to examine decision confidence in light of how often decision makers change their initial decision when they are offered a switching opportunity.

Furthermore, it is possible that both the quality and the size of consideration sets are affected by the use of preference relaxation. The size of the consideration set represents the number of alternatives that a decision maker is seriously considering. Thus, we plan to investigate the following hypotheses: Preference relaxation positively influences decision quality. Preference relaxation positively influences decision confidence. Preference relaxation leads to higher quality consideration sets. Evaluation To illustrate the potential benefits of the decision aid, we develop a use case based on the task of buying a used car using an online car market.

The dataset for our evaluation consists of real car advertisements from a dedicated online car-advertising site. Based on the study of relevant literature we identified two methods for evaluating our approach. Firstly, we propose to evaluate our propositions through a controlled user-based laboratory experiment.

The process of filtering involves application of filtering rules or restriction on attributes to the items in the set to be filtered [12]. Filtering rules are typically Boolean - if particular item does not satisfy any of the filtering rules it is removed from the set. Thus, users are able to significantly limit the number of alternatives they will examine in more detail with constructed filtering rules resembling their information preference. User preferences on attributes' values are the key input in the process of filtration If an alternative information item does not satisfy all the criteria specified by the user, it is removed from the set.

Thus, a new set is constructed, that contains only the alternatives that fully satisfy users attributes' values preference: Utility values of attributes As stated in the previous section, the model for overall value of the alternative requires a function for mapping the attributes' values and the user preferences into utility values. Utility evaluation is an important problem addressed by many researchers in different contexts and decision environments [5, 13, 14].

Intelligent product search with soft-boundary preference relaxation

In general, one can identify cost and benefit attributes in MADM problems [13]. Typically, the goal of a decision maker is to minimize the value of the cost attributes and maximize the value of the benefit attributes e. Linear normalized utility function for a cost-type a and benefit-type b attribute. In our model, the user's preference on attribute value can be expressed using an interval.

Thus, we define a simple utility mapping function as follows. For the convenience of calculation and simplicity, we propose a linear utility function see Fig. Thus, the linear utility function has the form of: Decision makers' preference of attribute values is p! Based in these assumptions and using 5 we calculate utilities of attributes' values in case of price utility 1 and size utility 2 of the apartment.

Soft-boundary preference decision aid One of the contributions of this research in progress paper is the soft-boundary filtration decision aid. We propose a decision aid that would limit the negative effects of the dynamics in the decision makers' preferences on attributes' weights and values. We present an example of application of the soft-boundary filtering decision aid in such a configuration in the following section. Illustrative example In the following example we use a multi-attribute decision-making problem of apartment selection adapted from [16] to illustrate the potential benefits of soft- boundary filtering decision aid.

Each apartment is described using the set of three following attributes: Size in square meters Among these attributes, g3 is of benefit type the higher value the better and gi and g2 are of cost type the lower value the better. Before examination of the initial set of alternatives see Table 2 , decision maker provides the importance level of attributes Weights and his preference on attributes' values Preferences see Table 3.

These information preferences allow the construction of the filtration criteria, the reduction of the set of alternatives so that only alternatives of value to the decision maker are presented see Table 4 and calculation of the utility values for price utility 1 , distance utility 2 and size utility 3 attributes see Table 5. However, user preferences are context dependent they can change significantly when a decision maker is comparing alternatives. We now examine a scenario where a soft-boundary filtration decision aid is used.

Thus, the decision maker is presented with the set of three x1; x2, x3 alternatives see Table 6 instead of two when standard filtration is used. This example illustrates how the use of a soft-boundary filtration decision aid can prevent decision maker from filtering out the alternatives he may find attractive from the initial set, and lead to more confidence in decisions in the decision process where decision maker's preferences are dynamic. Using the typical filtration, the decision maker eliminates all alternatives that do not fully fit his preferences specified using attributes' values ranges.

In contrast, when using the soft-boundary filtering decision aid, selected alternatives that do not fit initial decision maker's preferences are not removed from the set and can still be considered.

Search form

Our example illustrates how the variation of the decision maker's preferences may lead to choosing an alternative initially removed from consideration apartment x3 in our example. Moreover, we show that the soft- boundary filtering decision aid may lead to the increase of the quality of a decision. In this section we discussed the model of the problem under study and introduced the soft-boundary filtering decision aid. We argue that the soft-boundary filtering decision aid has significant impact on characteristics of the decision made. We continue discussion and elaborate the hypotheses to be tested, in the next section.

Research plan and hypotheses In the previous section we discussed the model of the soft- boundary filtering decision aid.

Intelligent product search with soft-boundary preference relaxation

In this section we debate the potential impact of such a decision aid on decision quality and consideration sets. Moreover based on the discussion in the previous sections we develop hypotheses to be tested in the future studies. Dependent variables We expect that the soft-boundary decision aid will have an impact on various aspects of multi-attribute decision- making, in particular: We discuss our expectations in more detail further in this section.

Decision quality is conceptualized as a decision maker's degree of confidence in a correctness of his decision discussed in the context of online shopping by [17]. We propose to measure of objective decision quality by how often decision makers change they initial decision when they are offered this opportunity.

Switching indicates low quality of the initial decision. More measures will be used when required by the experiment setup. Consideration set is defined as a set of alternatives that a decision maker seriously considers as his final choice. We argue that both the quality and the size of the set are affected by the use of the decision aid we propose.

The size of the consideration set is simply the number of alternatives that a decision maker is seriously considering. The quality of the consideration set can be assessed by the average overall utility value of alternative seriously considered by a decision maker. Decision Quality Research on decision support systems indicates that decision aids designed to screen large numbers of alternatives may reduce decision makers' cognitive effort [10] and improve decision quality by enabling individuals to make complex decisions with high accuracy [18].

Decision aids allow decision makers to significantly reduce the amount of unnecessary information processed. Moreover, the ability to screen alternatives in an efficient manner enhances the "quality" of the information that is processed, which, combined with reduced information quantity, should have a positive impact on decision quality [17]. Moreover, some results [19] suggest that "electronic decision formats based on weighted average scores for alternatives lead to less switching after initial choice".

Intelligent product search with soft-boundary preference relaxation

Thus, we argue that decision makers' use of the soft-boundary filtering decision aid will have the positive effects on the decision quality. In particular, the following hypotheses are to be tested: Use of the soft-boundary decision aid positively influences quality of the alternative being selected H2: Use of the soft-boundary filtration decision aid leads to a higher degree of confidence in the decision. Consideration Set Quality The soft-boundary filtering decision aid prevents filtering out the alternatives that do not fully satisfy the initial attributes' values preferences expressed by the decision maker.

Thus, it increases the probability that more alternative of potential high-value will be examined and perhaps considered by the decision maker. We expect that the average quality of the alternatives in the consideration set will increase when using the soft-boundary filtering decision aid. Use of the soft-boundary decision aid leads to the increase of the average overall utility value of the alternatives in the consideration set.

Consideration Set Size Model of consideration set size [2] suggest that the decision maker will continue to search for alternatives to consider "as long as the expected returns from search in terms of making a choice of higher expected utility exceed the cost of further searching". The soft-boundary filtering decision aid may have similar, however less intensive, effect on the size of a consideration set at it increases the number of alternatives considered by a decision maker.

Use of the soft-boundary decision aid leads to reduction number of items in consideration sets in comparison to no aid. Use of the soft-boundary decision aid prior to screening will increase the number of high-quality alternatives considered by the decision maker in comparison to no standard filtration decision aid.

Dataset for Experimentation In order to evaluate our research hypotheses, we plan to conduct a set of simulations and user based studies in various scenarios including housing market and used car market. For simulations in the used car domain we collected car advertisements that were posted and available for browsing.

Each advert was described using the following types of attributes: One image per advert was collected where available. The data allowed us to perform experiments in a controlled environment using an authentic dataset. Many decision aids e.

Thus, researchers often point out that the quality of the support provided by the various decision aids is highly dependant on the quality of data and is difficult to predict when dealing with real data.