Gambardella Flashcards
How to deal with strategic decision-making under uncertainty?
High-impact, low-frequency decisions such as investments in new technologies or top management recruitment require a theory-based approach, since there is no historical data or analogues to draw from.** The process is akin to the scientific method, relying on theories and experiments, combined with intuition and creativity.**
Which types of theories should we experiment?
The framework advocates for the experimental exploration of theories that, despite having lower expected value, might have higher variance, as they allow for greater learning. It is suggested that confirming a strongly believed theory is suboptimal compared to exploring more uncertain options.
What is the difference between Strategic management decisions under uncertainty and classical decision problems?
In classical decision problems, decision-makers calculate the expected values of different actions based on known states of the world and their probabilities. However, in strategic management, these states and probabilities are often unknown, leading to decision-making under state uncertainty. In such situations, decision-makers can use past data and experience to estimate the probabilities of various outcomes, as Luxottica did with its control over opticians in new national markets.
However, for completely new strategic decisions, both states and probabilities are unknown, and decision-makers must rely on theories that represent potential future states. These theories help to envision new, valuable state spaces that other firms have not yet tapped into.
What is the un-caused cause?
The un-caused cause is a starting point for the theory without further justification within the model.
These un-caused causes are important as they can either be a strong belief, a conceptual commitment to a hypothetical action, or an assumption about an exogenously determined future state.
Example: The term P(Y), which represents the probability of Luxottica being able to design, produce, and market the eyeglasses effectively, is identified as an “un-caused cause”
What is the difference between a Model and a Theory?
A model is a specific probability distribution over the state space defined by attributes and causal links.
A theory is a set of all probability models compatible with the causal links and beliefs of the decision-maker. A theory thus restricts the choice of parameters to a subset compatible with the decision-maker’s beliefs.
Example: Luxottica’s theory is that the ability to design, manufacture and market fashion glasses raises the probability of high demand. Thus, θHN and θY can take any value, but θHY must be such that θHY > θHN
What is the Dirichlet Distribution and its main assumption regarding states?
Dirichlet Distribution: A Dirichlet distribution is a type of probability distribution that is often used to model the probabilities of outcomes in a multinomial process. Parameters
ηIJ greater than zero correspond to each of the four states (High demand with successful design H,Y, High demand with unsuccessful design H,N, Low demand with successful design L,Y, Low demand with unsuccessful design (L,N)
It is assumed that the probabilities of the four states in the previous section are jointly distributed as a Dirichlet distribution
Expected Probability: The expected probability of each state is calculated by dividing the parameter of interest (ηIJ) by the sum of all parameters. This gives the mean estimate of the probability for that state in the Dirichlet distribution.
How can we obtain the Expected Value of the Theory?
This is a more complex concept that involves integrating over all possible models within the theory to get an average or expected outcome. It takes into account all the beliefs and uncertainties that the theory encompasses.
How is the unconditional expected value of the theory calculate?
is a weighted average of the expected value assuming the theory is true (Vo) and the expected value under the null hypothesis, with weights given by the decision-makers’ confidence in the theory (w) and its complement (1-w):
V = wVo + (1 - w)V~
What is the effect of an experiment?
Explain what it does change
Decision-makers conduct experiments as deliberate attempts to collect evidence about a phenomenon, helping to update their theories.
* Updating Distributions: Through experiments, they update the probability distribution of parameters (theta) in their models, which consequently also changes the conditional probability distributions and the expected values Vo. V~, and the unconditional V.
Conditional Expected Value: After an experiment, the conditional expected value of the theory, denoted by V, changes based on the updated probability distribution (mu’). This value is obtained by integrating the utility function v(theta over all possible values of theta, weighted by the updated probability distribution mu’(theta.
Belief Update: If experimental evidence leads to a significant change in the expected parameters beyond a certain threshold, decision-makers will update their belief about the theory (w). When beliefs about the theory change, it affects the expected values Ve, W, and V, which are based on these beliefs.
When shuld we run an experiment of our main theory against an alternative theory?
Decision-makers should run an experiment on their current theory against an alternative theory if the net expected value of the experiment is greater than the cost of the experiment.
Net Expected Value: The net expected value is the sum of the expected value of the theory after updating it with new information from the experiment and the value of the alternative theory (Q), minus the cost of the experiment.
THUS, The experiment is worth running when its net value V E is higher than the highest current probability of success.
What is the difference between Falsification and Confirmation Experiments?
Falsification Experiments: These are tests designed to challenge the dominant theory. If new information indicates a potential opportunity that is more valuable than what was previously believed, a falsification experiment could lead to the adoption of a new, innovative theory.
Example: Luxottica tested the new theory of eyewear as a fashion item, which was less conventional and had more uncertainty. This experiment was considered a falsification experiment because it challenged the existing business model.
Confirmation Experiments: These are tests conducted to confirm the current theory. These may be done when external shocks or new information suggest that the current business model may be less valuable than previously thought.
Example: Had Del Vecchio considered the rise of eye surgery as a more uncertain theory, he could have opted for a confirmation experiment to test whether his current business would remain viable in the face of this potential disruption.
