Archive for the ‘Game Theory’ Category

Judgment under Uncertainty – Heuristics and Biases (Kahneman & Tversky, 1974)

Dienstag, August 12th, 2008

Judgment Heuristics and Biases

Tversky, Amos; Kahneman, Daniel: Judgment under Uncertainty – Heuristics and Biases; in: Science, Vol. 185 (1974), No. 4157, pp. 1124 – 1131.
DOI: 10.1126/science.185.4157.1124

Biases have evolved to lower our energy needed to make decisions, so they do have quite a natural place in our ape-sized world. Last time I checked wikipedia lists 100 biases, heuristics, and memory errors. Kahneman & Tversky published the first theorization in this article [also published as a part of this book].

Starting with the now classical example of the Gambler’s fallacy the authors explore three judgment heuristics commonly found in science and economic decision making (1) Adjustment & Anchoring, (2) Representativeness, and (3) Availability.

Anchoring & Adjustment (Decisions often rely on a single piece of information) – Kahneman & Tversky show that persons usually guess probabilities more accurately if they have been presented with an anchor. They show that students do overestimate their success when asked at the beginning of a term. This overestimation is slightly corrected if they were given or asked for an anchor, such as ‚what do you think was the grade distribution of your fellow students last term?‘.

Representativeness (Commonality is assumed for similar events or objects) – The authors describe several misconceptions of chance and insensitivities to prior probabilities, sample sizes, and predictability. They also describe the illusion of validity, but the the misconception of regression is the most important of these biases. It is also the reason why we have control groups in double-blind experimental studies.
Regression towards the mean means that in any given random process every sub-group will produce the same distribution [give or take effects of the sample size]. For example, assume that a group has been split into quartiles according to the results after the first run of the random process. The repetition of this process will automatically produce the same distribution in each sub-group, thus the bottom quartile will be better and the top quartile will perform much worse without any effect of a stimuli which has been applied.

Availability (Expected probabilities influenced by the ease of brining examples to mind) – In their classical example for the retrievability bias subjects have been asked to estimate the proportion of words in the English language that start with R or K and the proportion of words that have R or K as a third letter. This bias leads people to underestimate the number of words with R or K as a third letter.

The impact of principal–agent relationship and contract type on communication between project owner and manager (Müller & Turner, 2005)

Montag, Juli 14th, 2008

Communication and Contract type

Müller, Ralf; Turner, Rodney J.: The impact of principal–agent relationship and contract type on communication between project owner and manager; in: International Journal of Project Management, Vol. 23 (2005), No. 5, pp. 398-403.
http://dx.doi.org/10.1016/j.ijproman.2005.03.001

Müller & Turner argue that a project should be analysed in two distinctive stages pre and post contract. The pre contract perspective follows the laws outlined in Transaction Economics (or Transaction Cost Theory), which postulates that companies choose the governance structure with the lowest transactions costs. Specifically the degree of asset specificity, the degree of uncertainty, and the frequency of transactions determine the transaction costs both parties (vendor and buyer) consider in this phase.

The post-contract perspective is best conceptualised by the Principal-Agent-Theory, specifically by with the phenomena of moral hazard and adverse selection. Game theory usually calls this equilibria with correlated strategies and calls for contractual design in a way that each player is receiving incentives for compliance.

On a side-note: This behaviour is a classical chicken game. Imagine a vendor and a buyer who build up an integrated team. Both agree (orally) to send their most experienced experts (unfortunately they cost more), which would be best for the project work to be done. In this game each party can choose to give the work either to experts or to rookies:

  Rookies Experts
Rookies (0,0) (8,2)
Experts (2,8) (6,6)

Thus it is better for each team to sent rookies if the other team sends experts (which gives us the two Nash-equilibria of the game). If vendor and buyer send their rookies no one will know what is going on and what they have to do and both earn nothing (0,0). On the other hand the oral agreement of both parties to send their experts is not credible. Since this is a one-shot game the strategy (6,6) can only be reached if both parties agree on binding contracts or a negotiator.

Müller & Turner show that decisions made in the pre-contract phase with the objective to minimize the transaction costs (e.g. fixed-price or cost-plus contracts) can potentially have
have adverse effects on the project in the post-contract phase – by increasing the administrative costs for communication due to their negative impact on owner–manager
collaboration.