# Lecture 019

Stance: neither of the two extreme (pure description vs. universal law), middle range theory

Unify tree well-know theories (by causal modeling)

• self-fulfilling

• network diffusion

• threshold-based behavior

Mechanism: middle ground between pure description and universal law

• Jon Elster: flexible level research

• Negative Definitions

• causal explanation != causal statements. need cite causal mechanism

• causal explanation != correlations
• causal explanation != necessitation
• causal explanation != story-telling
• causal explanation != prediction (we can explain without being able to predict, predict without being able to explain) if we predict without explain, will it be causal?

Stinchcombe: mechanism is a reduction of a theory to lower (individual) level, the purpose is to make higher-level more accurate or more general

Explanatory mechanism: can explain things that statistics (correlation) cannot

Hempel:

• social law are probabilistic

• social law does not reveal internal mechanism (only "black-box" prediction)

benefits of generative mechanism

• distinguish between causality and association

• increase understanding of why we observe what we observe

Blackbox

• regression coefficient did not consider what's in the black box

• behaviorist only talks about correlation, so it is black box. On the contrary, rational-choice-theory introduce preference and intention, which is not a black box.

Generality of a law makes a law convincing (no ad hoc)

Unobserved assumption: can be only tested by confirmation

• making hypothetical model is a good way to move science forward

• Hauser: statistical data is explanation

• Boudon: statistical data is not mechanism

## Questions

1. above example of theory generalizable to make them "mechanism"? "bandwagon effect" "production" "consumption"

Bandwagon effects tries to explain behavior by stating that "people have tendency to imitate other's behavior" (page 284). The "production" and "consumption" theory (page 286) in microeconomics tries to model the price of objects in market. Because these theory can be generalized to some extent in their fields of study, but they are not universal laws that can't be violated, they can be considered "middle-ground explanation," and therefore these laws satisfy the "generalizability" of "mechanisms." (For example, bandwagon effects can be used to explain almost any behavior since human behaviors overlap significantly. Production and consumption theory can be used in any market for any products.) However, because, to my knowledge, bandwagon effects does not try to explain why "people have tendency to imitate other's behavior," it is also considered a "black-box explanation" that does not make it a "generative mechanism" (page 287). Therefore, bandwagon effects offers more as a "story-telling" and "prediction" than a "causal explanation" (page 283).

1. only moving association to finer level? Yes. but as we move finer and finer level, the more certain we are about causal chain

2. mechanism better than correlation? Mechanism is better, cuz mechanism adds info to correlation

4.

## Class

### Types of acceptable mechanisms

Situational mechanism: how social level conditions affect agents (social affect individual) Individual action mechanism: how desires/beliefs/opportunity-to-act generates a specific individual action (psychology affect individual) Transformations mechanism: how a number of agents, individually acting in interaction with one another, produces collective outcome (prisoner's dilemma)

Table of Content