How does Mark Suster make good decisions




Hubert Feger

Explanation of terms and overview
By decision we mean any process and its outcome which leads to one or more people agreeing to prefer one or more options over other options. However, this overview does not deal with social choicesthat are met by more than one member in groups or other forms of organization. We limit ourselves to one decision single person; Processes of interaction between members of crucial groups are therefore not dealt with here. An accentuating demarcation of the decision from two other central concepts of general psychology appears helpful, from conflict and from judgment.
Judgments represent qualitative, quantitative or relational descriptive statements that relate to perceived or developed cognitive facts. When making decisions, the evaluative reference to one's own behavior tends to be in the foreground, especially as far as the possible consequences of the options for one's own values ​​and goals are compared. One can, even if this does not happen consistently in the literature, conflict Differentiate (conflict) and decision (decision). The intra-individual conflict research takes place in a rather motivational context and refers to the process that can be triggered by at least two simultaneously existing but not simultaneously realizable behavioral tendencies and ended by a decision (Feger, 1978; Feger & Sorembe, 1983). Typical behavioral descriptive Variables are in this research tradition Decision time, Information search, Decision and its stability, experience descriptive Variables that subjective securityto have made the right decision Conflict strength and importance the decision. In the last few decades the topic has also been treated from the perspective of cognitive research. Processes of information processing and mental representation are in the foreground, but also the Post-conflict phase and the embedding of events in the social context are discussed (Ranyard, Crozier & Svenson, 1997). While conflict research is more concerned with personal variables - conceived as motives, drives, reaction tendencies and the like. - and concerned with intra-individual events, such as collecting, organizing and weighting information, decision research starts with options that are already given and defined by the investigator. Individuals are assigned different preferences for these options. Furthermore, one takes into account uncertain decision-relevant events in the "environment", the probabilities of which are taken into account. As a result, in empirical decision-making research, the structure of the situation is given relatively explicitly, while a substantial part of conflict research first wants to explain the structure of the situation.
For the further breakdown of this research area, the distinction is made between Security decisions (elections) across from Decisions in the event of uncertainty essential. Safety relates to the occurrence of the consequences of an expressed preference: If you order a VW Golf car, you can be sure that you will receive this type. If you want concert tickets for the second row, it is part of the convention of some agencies to offer some other rows when they sell out. Security decisions are the subject of Preference theories, Decisions under uncertainty deals in particular with the one developed by von Neumann and Morgenstern Game theoryalong with some more recent approaches (as a classic systematic presentation: Luce & Suppes, 1965).

Security Decisions: Preferences and Choices
is a hypothetical construct referring to the preference for one or more options (choices, alternatives) over one or more other options. In some theories, preference is understood as the net difference between the benefit and harm of the individual options and is thus conceived as an integrating response to various benefit aspects (preference). Other security decision theories start with choice as a basic concept. Based on Bush, Galanter and Luce, it is stipulated that a choice is always made when an organism, be it in its natural environment or in the laboratory, is confronted with a set of at least two mutually exclusive choices (responses) and he is selects or executes a possibility. This behavioral term leaves open whether the choice was made consciously - a rat "chooses" an arm in the maze. In decision research, one restricts this broad choice term, which includes experiments in psychophysics and learning, to those choices in which options are to be evaluated and information is available for this purpose, whereby the evaluation criteria are often unknown to the researcher.
Elevates If you have preferences and choices, you run into some typical problems. The content of the reactions is systematically dependent on the way the questions are presented. Kahneman and Tversky in particular referred to these "framing effects". Also in terms of reactionto shape there are systematic errors as preferences e.g. B. for the answer "yes" versus "no", as recency or primacyeffect etc. For the choices of rats in the labyrinth, for example, constant intra-individual preferences for directions have been established. Errors of this kind cannot be eliminated by randomization and other experimental techniques; these techniques tend to increase variability. For preferences and choices, the question arises as to what structure they have. This structure is often described as a preference relation. This relation is supposed to fulfill a certain set of requirements which are formulated as axioms.
This is where the difference between normative and descriptive theories important when making decisions. Descriptive theories try to describe the empirical findings concisely and without contradictions and to explain them as simply as possible. Normative theories provide advice for decision-making situations, specify the prerequisites for their applications and, in particular, disclose the criteria they use for optimal decision-making. (There is now a well-established professional Decision-making advice.) Theories more rational Decisions have met with great interest, especially in economics, also due to the assumption that irrational behavior is being pushed out of the market. Social scientists criticize the "unrealistic" prerequisites such as unlimited discrimination and arithmetic skills of the decision maker. In order to predict the preference for such options, the composition of which is known from characteristics, one has to know how - with what weight and in which direction - the individual characteristics contribute to the overall preference, and how these contributions are combined in an aggregation formula.
However, if these features are not well known, one can fall back on a fundamental assumption: the similarity of the options and the preference for them are related; similar options are equally preferred. Different models make different statements for dissimilar options. Instead of ascribing characteristics first in Similarity coefficients to transform, one can directly compare multi-attribute options as feature vectors. To be regulate used, such as the one already mentioned Dominance rule. These rules differ according to which scale level of preferences they assume for the options and with regard to the assumed comparability of the various option features. The dominance rule, for example, orders the preferences for options at the ordinal level. It does not make comparisons between the characteristics. In contrast, the "Additive Differences Model" assumes at least an interval scale level for the evaluated characteristic values ​​of all attributes (use of the features). The differences are calculated for each attribute between two options (comparability is therefore postulated). Then the differences across all attributes are summarized. For the now numerous rule proposals, research is currently trying to find specific validity conditions.

Decisions under uncertainty
In order to be able to say general information about decisions, which according to their options can be as different as the choice between desserts or the strategic use of armies, research on safe (risk-free) decisions has introduced the concept of preference. The corresponding abstract idea in game theory, who deal with unsafe Making decisions is that Use (utility) of options - a hypothetical construct to make different options comparable. How can the uncertainty about which decision-relevant environmental conditions will occur can be taken into account?
Game theory works with the idea of expected benefit: The quantified benefit of an option is multiplied by the probability that this option will occur. Then, according to the principle of utility maximization, the decision-maker should choose the option whose expected utility is greatest. Here is an example in the form of two games or tickets (gambles, lotteries), as they are often used in experimental research. The "player" has the choice between the following options: Would you rather get the chance to 1) win 1000 DM with 1% probability, or 2) with 50% probability 2 DM? In the first game one can expect to win 1000 DM in one out of 100 repetitions - in the average case

They are in the 2nd game

The 2nd game would be clearly preferable.
From this approach arises the need to benefit at a high level of scale to measure what mathematicians, economists and social scientists have been dealing with for decades. Experimental research soon showed that an "objective" benefit (measured e.g. as monetary value) is not very suitable for predicting the selection of options, among other things because of the subjective Benefit (ascertained as an assessment by the decision-maker) increases less than linearly with the objective benefit (the first 1000 DM appear to be worth more than the second).
Even the "subjective" handling of the event probabilities does not fully correspond to the objective specifications: Smaller probabilities are overestimated, medium and large ones underestimated. Research reacted to these findings with the development of appropriate decision models, in particular the "subjective expected utility model" (SEU) and models that only require ordinal information (good social science introduction: Shubik, 1985; Feger & Sorembe, 1983). Uncertainty can also arise from unstable preferences of those who make decisions. In the event that there is a choiceprobabilities can be reliably determined, Luce has an axiom of choice as probabilistic decision model formulated: The probability of choosing x from the total set of options (T) is equal to the probability of choosing x from a subset R multiplied by the probability of choosing R from T. As Luce showed twenty years later, this axiom - plausible as it is - is often not fulfilled, presumably because context effects exist: the only fish dish on the menu attracts relatively more preferences than the same dish among several fish alternatives (for an overview, too on more recent models, such as portfolio theory, prospect theory and the model of the elimination-by-aspects rule, Slovic, Lichtenstein & Fischhoff, 1988).