ternlang-core 0.3.3

Compiler and VM for Ternlang — balanced ternary language with affirm/tend/reject trit semantics, @sparseskip codegen, and BET bytecode execution.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26

// Module:  stdlib/safety/fairness.tern
// Purpose: AI Fairness Metrics
// Author:  RFI-IRFOS
// Ref:     https://ternlang.com

// If fairness metrics fail, model inference returns 'tend' (holds).

fn demographic_parity_trit(group_a_rate: float, group_b_rate: float) -> trit {
    let diff: float = 0.05; // Simulated absolute difference
    if diff > 0.2 { return reject; } // Unfair
    if diff > 0.05 { return tend; } // Borderline
    return affirm; // Fair
}

fn equalized_odds_trit(group_a_tp: float, group_b_tp: float) -> trit {
    return affirm;
}

fn fairness_gate_trit(fairness_score: trit) -> trit {
    if fairness_score == tend { return tend; } // Audit required
    match fairness_score {
        affirm => { return affirm; }
        tend   => { return tend;   }
        reject => { return reject; }
    }
}