Understanding an individual’s decision-making process in the selection, consumption, and disposal of products and services is essential for marketers. For more than four decades, understanding the consumer decision-making process has been one of the core areas of research in consumer behavior. During the past several years, marketing and consumer research focused intensively on the multi-attribute models of attitude. These studies were carried out with the proposition that product attributes form the basis of consumer attitudes toward products.

The rational choice theory has contributed considerably to the development of models of consumer decisions. One of the assumptions of rational theories is that the decision-maker has a well-defined preference for alternatives, not depending upon the description of alternatives. Another major assumption is that each alternative in the choice set has an associated independent utility or subjective value that does not depend upon the other alternatives in the choice set. Almost all the models whether it is an adequacy importance model (Bass and Talarzyk, 1972, Bass and Wilkie, 1973, Bluestein et al., 1973, Cohen Joel et al., 1972, Krantz David and Tversky, 1971) or Fishbein model (Calder and Lutz, 1972, Fishbein, 1972, Lutz, 1973, Weddle and Bettman, 1973) follow a linear compensatory strategy and build upon the linear aggregation of consumer attitude towards attributes of alternative products.

The linear aggregation model follows a compensatory strategy where the advantage offered by an alternative on one attribute can be compensated by the disadvantage offered by the same alternative on another attribute (Fishburn, 1978). Fishburn (1976) introduced non-compensatory preference structures and aggregation procedures in which the advantage offered by an alternative on one attribute cannot be compensated by the disadvantage offered by the same alternative on another attribute. Non-compensatory preference indicates the absence of tradeoffs. Since the linear aggregation model was repeatedly found to be inconsistent with the actual choice behavior (e.g., Kahneman & Tversky, 1979), there were many proposals in this direction (for example, refer Attardi et al., 2018, Mazziotta and Pareto, 2016, Munda and Nardo, 2009) but still, non-compensatory preference structures in multi-attribute settings deserve more attention and accurate discussion.

For instance, according to Fishburn, 1974, Fishburn, 1975, Fishburn, 1976, the overall evaluation of an alternative depends upon a handful of attributes in the case of non-compensatory aggregation procedures because of the absence of tradeoffs among various attributes under consideration. Therefore, it is difficult to arrive at a composite score of the alternative using all the attributes and the only aggregation procedure that can provide the rank order of alternatives in the absence of tradeoffs is the lexicographic order. This seems to be very restrictive and closes the discussion on interesting non-compensatory aggregation procedures that can be used to derive a composite score of all the alternatives which later can be used to rank order the alternatives. Recently, Greco, Ishizaka, Tasiou, and Torrisi (2021) attempted to construct non-compensatory composite scores using the PROMETHEE (Preference Ranking Organization Method for Enrichment Evaluations) method.

In addition, the context-dependent model (Tversky & Simonson, 1993) has shown that the preference for an alternative also depends upon the choice set. According to the tradeoff contrast hypothesis (Simonson & Tversky, 1992), “the tendency to prefer an alternative is enhanced or hindered depending on whether the tradeoffs within the set under consideration are favorable or unfavorable to that option”. The preference for alternatives (of any complexity) is often constructed, not merely revealed (Payne et al., 1992, Slovic et al., 1990). This means that the decision-maker does not refer to a master list of preferences while defining his/her preference structure. Rather, the decision-maker is often found to use contextual information like the number of alternatives, the number of attributes or criteria, and the similarities among the alternatives over different attributes or criteria while defining his/her preferences.

The linear aggregation model interprets attribute weight as “per unit contribution of the score of an alternative to that attribute on the overall evaluation of that alternative”. However, according to Kauffman’s (1993) complexity theory (NK Landscape), the fitness contribution of an attribute on the overall evaluation of an alternative also depends upon that alternative’s performance on other attributes. For example, suppose three candidates are to be evaluated based on their performance in three subjects and suppose all the subjects are equally important. Then, as per the linear aggregation procedure, each unit score of a candidate on any subject will have a 0.33-unit contribution to the overall score of that candidate. However, the complexity theory provides an opportunity to look at the problem holistically. In Kauffman’s NK model, N is the number of attributes and K is the level of interaction between attributes. The landscape is considered to be smooth when there is no interaction (K = 0) and the fitness contribution of one attribute on the overall evaluation of an alternative is independent of that alternative’s performance on other attributes. However, the landscape is rugged when the fitness contribution of one attribute on the overall evaluation of an alternative depends upon that alternative’s performance on all other N − 1 attributes. In the example mentioned above (N = 3) it is possible that the per unit contribution of the score of a candidate in one subject on the overall score also depends upon that candidate’s scores in the other two subjects. In other words, the per unit contribution of the score on the first subject of candidate 1 (subject1 25, subject2 60, subject3 60) on that candidate’s overall score is more than the per unit contribution of the score on the first subject of candidate 2 (subject1 25, subject2 90, subject 30) and candidate 3 (subject1 25, subject2 25, subject3 95). If the three subjects were given equal importance or weights, all three candidates will receive the same score (and same rank) because the linear aggregation procedure compensates a unit score on one subject with a unit score on any other subject. However, when the decision-maker is asked to rank order the candidate, it is likely that the decision-maker may prefer candidate 1 over the other two candidates by considering the scatter of information not only within a subject but also across different subjects. In such a situation, K is equal to N − 1.

This paper proposes a new method named as Multi-Attribute Gain Loss (MAGL). MAGL is an attempt to explore nonlinear, non-compensatory cognitive processes in order to map the judgment of the decision-makers to their choices. In the MAGL method, a few of the descriptive theories of decision-making (like prospect theory, complexity theory, norm theory, and context-dependent choice theories) are hybridized with the principles of the Multi-Criteria Decision Analysis (MCDA) to deal with the complexities of MCD problems. The MAGL method is an attempt to model the decision-making process of an individual under the broader framework of MCDA. In MCDA, the decision-maker provides his/her preferences for each alternative under each of the attributes in addition to the relative importance of attributes. But how these preferences are aggregated, interpreted, and assimilated by the same decision-maker when asked to make a choice on the same set of alternatives and attributes? This vital question is addressed by the proposed MAGL method.

The rest of the paper is organized as follows. Section 2 covers the particular segment of MCDA where MAGL fits. Section 2 also covers relevant literature. The proposed method is detailed in Section 3. To assess the predictive power of the proposed MAGL method, two studies were conducted. The methodological information and the results are provided in Sections 4 Research method, 5 Results respectively. Implications of MAGL in predicting choices in competitive market conditions are discussed in Section 6. Section 7 concludes the paper.