Struct Explanation 

pub struct Explanation {
    pub id: u64,
    pub style: usize,
    pub style_name: String,
    pub utility: f64,
    pub utility_std: f64,
    pub mix_utility: f64,
    pub responsibility: f64,
    pub contributions: Vec<Contribution>,
}

Why the model scores one candidate the way it does.

Utility is exactly linear within a style lens, so this decomposition is exact rather than a local surrogate: Σ contribution = utility. No SHAP, no LIME, no approximation error to caveat.

Fields

id: u64

Candidate id.

style: usize

Aligned index of the lens that claims this candidate.

style_name: String

That lens’s user-given name ("" if unnamed).

utility: f64

Posterior-mean utility under that lens — exactly the sum of the contributions. This is the quantity the decomposition explains.

utility_std: f64

Posterior std of that same lens utility — how sure the model is about this score.

mix_utility: f64

Posterior-mean mixture utility E[max_k u_k] — the number the bank is ranked by, and the one to show as the score.

It is not the same number as utility, and it is never smaller: the ranking takes the max over lenses inside the expectation, while the decomposition necessarily fixes one lens first. Jensen’s inequality does the rest. Showing utility next to a bank ordered by mix_utility would render a systematically lower number beside the row it is supposed to explain.

responsibility: f64

Posterior probability that style really is this candidate’s best lens. Near 1, utility ≈ mix_utility and the explanation is the whole story; well below 1, the candidate sits between islands and the gap is worth surfacing rather than hiding.

contributions: Vec<Contribution>

Every feature’s contribution, sorted by descending magnitude.

Trait Implementations

impl Clone for Explanation

fn clone(&self) -> Explanation

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Explanation

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for Explanation

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for Explanation

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.