SOLUTION: Suppose that Alice and Bob have the following credence functions defined over the propositions p, q, and p∨q: P Q P ∨ Q Alice .8 .2 .9 Bob .1 .3 .4 Suppose that p is true

Algebra ->  Probability-and-statistics -> SOLUTION: Suppose that Alice and Bob have the following credence functions defined over the propositions p, q, and p∨q: P Q P ∨ Q Alice .8 .2 .9 Bob .1 .3 .4 Suppose that p is true       Log On


   



Question 1160759: Suppose that Alice and Bob have the following credence functions defined over
the propositions p, q, and p∨q:
P Q P ∨ Q
Alice .8 .2 .9
Bob .1 .3 .4
Suppose that p is true and q is false. Compute the accuracy of Alice and Bob’s
credences using the Brier score. Which of the two is more accurate?

Answer by CPhill(2303) About Me  (Show Source):
You can put this solution on YOUR website!
```python
# Given:
# Truth values:
# p is True -> v(p) = 1
# q is False -> v(q) = 0
# p v q is True (since p is True) -> v(p v q) = 1
# Credences:
# Alice: c(p) = 0.8, c(q) = 0.2, c(p v q) = 0.9
# Bob: c(p) = 0.1, c(q) = 0.3, c(p v q) = 0.4
# Brier score definition in formal epistemology:
# The Brier score (squared error loss / penalty score, or accuracy score).
# Standard Brier score (penalty/inaccuracy):
# BS(c, v) = sum_{X} (c(X) - v(X))^2
# Sometimes defined as mean squared error: (1/|Propositions|) * sum (c(X) - v(X))^2
# Or accuracy score: 1 - sum or -sum.
# Let's calculate both sum of squared differences and mean squared differences.
v = {'p': 1, 'q': 0, 'pvq': 1}
alice = {'p': 0.8, 'q': 0.2, 'pvq': 0.9}
bob = {'p': 0.1, 'q': 0.3, 'pvq': 0.4}
# Sum of squared errors (standard Brier inaccuracy score)
bs_alice_terms = {k: (alice[k] - v[k])**2 for k in v}
bs_bob_terms = {k: (bob[k] - v[k])**2 for k in v}
bs_alice_sum = sum(bs_alice_terms.values())
bs_bob_sum = sum(bs_bob_terms.values())
bs_alice_mean = bs_alice_sum / len(v)
bs_bob_mean = bs_bob_sum / len(v)
print("Alice terms:", bs_alice_terms)
print("Alice sum:", bs_alice_sum, "mean:", bs_alice_mean)
print("Bob terms:", bs_bob_terms)
print("Bob sum:", bs_bob_sum, "mean:", bs_bob_mean)

```
```text
Alice terms: {'p': 0.03999999999999998, 'q': 0.04000000000000001, 'pvq': 0.009999999999999995}
Alice sum: 0.08999999999999998 mean: 0.029999999999999995
Bob terms: {'p': 0.81, 'q': 0.09, 'pvq': 0.36}
Bob sum: 1.26 mean: 0.42

```
### 1. Truth Values of the Propositions
Given that $p$ is true and $q$ is false:
* $v(p) = 1$
* $v(q) = 0$
* $v(p \lor q) = 1$ (since $p$ is true)
---
### 2. Brier Score Calculation
In formal epistemology and scoring rule theory, the **Brier score** measures inaccuracy as the squared difference between the credence $c(X)$ assigned to a proposition and its actual truth value $v(X) \in \{0, 1\}$:
$$BS(c, v) = \sum_{X} (c(X) - v(X))^2$$
*(A lower Brier score represents lower error, and therefore **greater accuracy**).*
#### Alice's Brier Score
* For $p$: $(0.8 - 1)^2 = (-0.2)^2 = 0.04$
* For $q$: $(0.2 - 0)^2 = (0.2)^2 = 0.04$
* For $p \lor q$: $(0.9 - 1)^2 = (-0.1)^2 = 0.01$
$$\text{Total Brier Score (Alice)} = 0.04 + 0.04 + 0.01 = \mathbf{0.09}$$
$$\text{Mean Brier Score (Alice)} = \frac{0.09}{3} = \mathbf{0.03}$$
---
#### Bob's Brier Score
* For $p$: $(0.1 - 1)^2 = (-0.9)^2 = 0.81$
* For $q$: $(0.3 - 0)^2 = (0.3)^2 = 0.09$
* For $p \lor q$: $(0.4 - 1)^2 = (-0.6)^2 = 0.36$
$$\text{Total Brier Score (Bob)} = 0.81 + 0.09 + 0.36 = \mathbf{1.26}$$
$$\text{Mean Brier Score (Bob)} = \frac{1.26}{3} = \mathbf{0.42}$$
---
### Verdict
**Alice is significantly more accurate than Bob.**
Her total Brier inaccuracy is **0.09** compared to Bob's **1.26** (or **0.03** vs. **0.42** if taken as the mean Brier score). Alice assigns high credence to the two true propositions and low credence to the false proposition, whereas Bob's credences diverge heavily from the truth.