deep_causality_ethos 0.2.6

Programmable ethics for DeepCausality.
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
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146

# DeepCausality Ethos

[![Crates.io][crates-badge]][crates-url]
[![MIT licensed][mit-badge]][mit-url]
[![Build Status][actions-badge]][actions-url]

[crates-badge]: https://img.shields.io/crates/v/deep_causality.svg
[crates-url]: https://crates.io/crates/deep_causality_ethos
[mit-badge]: https://img.shields.io/badge/license-MIT-blue.svg
[mit-url]: https://github.com/deepcausality/deep_causality.rs/blob/main/LICENSE
[actions-badge]: https://github.com/deepcausality/deep_causality.rs/workflows/CI/badge.svg
[actions-url]: https://github.com/deepcausality/deep_causality.rs/actions

**DeepCausality Ethos** is a programmable deontic reasoning layer for the DeepCausality stack. It evaluates a `ProposedAction` against a set of norms and returns a justified `Verdict` (`Obligatory`, `Impermissible`, or `Optional(cost)`).

The crate implements the teleological layer described in section 8 of the Effect Propagation Process paper. It pairs a defeasible deontic logic with the DeepCausality `Context` so that norms can read the same spatio-temporal state that causal reasoning operates on.

## Overview

The unit of regulation is a `Teloid`: a single norm that names an action, an activation predicate over `(Context, ProposedAction)`, a modality, and three heuristics used for conflict resolution (specificity, priority, timestamp). Teloids are kept in a `TeloidStore`, indexed by tag in a `TagIndex`, and linked in a `TeloidGraph` whose edges carry a `TeloidRelation` of either `Inherits` or `Defeats`.

The `EffectEthos` struct owns these components and exposes the reasoning API. Evaluation runs in five steps:

1. Tag-based filtering selects candidate norms from the `TagIndex`.
2. Each candidate's activation predicate is run against the `Context` and the `ProposedAction`. Uncertain predicates are tested with the teloid's `UncertainParameter` (threshold, confidence, epsilon, sample bound).
3. Active norms are reduced through the `Defeats` edges in the graph (defeasance).
4. Survivors are checked for consistency under `Lex Specialis`, `Lex Superior`, and `Lex Posterior`.
5. A `Verdict` is returned, carrying the final modality and the IDs of the norms that justify it.

The graph must be frozen and verified for acyclicity before evaluation; calling `verify_graph()` performs both.

## Features

* **Deterministic and uncertain norms:** `add_deterministic_norm` takes a `fn` predicate. `add_uncertain_norm` takes an `UncertainActivationPredicate` and an `UncertainParameter`, lifting probabilistic activation into the deontic layer.
* **Explicit conflict resolution:** specificity, priority, and recency are first-class fields on every `Teloid`. Resolution is deterministic and reproducible.
* **Auditable verdicts:** every `Verdict` carries a `justification: Vec<TeloidID>` so a decision can be traced back to the norms that produced it. The `DeonticExplainable` trait exposes this trail.
* **Context-aware predicates:** norms read the full DeepCausality `Context<D, S, T, ST, SYM, VS, VT>`, so deontic rules can depend on space, time, symbolic state, and data in one expression.
* **Static dispatch:** no `dyn` in the public API; the engine is generic over the same seven type parameters as the rest of the DeepCausality core.

## Public API

`lib.rs` exports:

* Types: `EffectEthos`, `Teloid`, `TeloidStore`, `TeloidGraph`, `TagIndex`, `TeloidModal`, `TeloidRelation`, `Verdict`.
* Traits: `DeonticInferable`, `DeonticExplainable`, `TeloidStorable`, `Teloidable`.
* Aliases: `BaseTeloidStore`, `TeloidID` (`u64`), `TeloidTag`.
* Errors: `DeonticError`.

## Usage

Add this to your `Cargo.toml`:

```toml
[dependencies]
deep_causality_ethos = "0.2"
deep_causality = "0.13"
```

### Building an EffectEthos

```rust
use deep_causality_ethos::{EffectEthos, TeloidModal, DeonticInferable};
use deep_causality::{ActionParameterValue, Context, ProposedAction};
use std::collections::HashMap;

// Define a deterministic predicate over Context and ProposedAction.
// "A drone must not take off when battery is below 20%."
fn battery_below_minimum<D, S, T, ST, SYM, VS, VT>(
    _ctx: &Context<D, S, T, ST, SYM, VS, VT>,
    action: &ProposedAction,
) -> bool {
    match action.parameters().get("battery_pct") {
        Some(ActionParameterValue::Number(pct)) => *pct < 20.0,
        _ => false,
    }
}

// Build the ethos with a single norm, then freeze and verify the graph.
let mut ethos = EffectEthos::new()
    .add_deterministic_norm(
        1,                            // TeloidID
        "takeoff",                    // action identifier
        &["flight_safety".to_string()], // tags
        battery_below_minimum,        // predicate
        TeloidModal::Impermissible,   // modality
        0,                            // timestamp
        10,                           // specificity
        100,                          // priority
    )
    .expect("failed to add norm");

ethos.verify_graph().expect("graph must be acyclic");
```

### Evaluating a proposed action

```rust
let mut params = HashMap::new();
params.insert("battery_pct".to_string(), ActionParameterValue::Number(12.5));
let action = ProposedAction::new(1, "takeoff".to_string(), params);
let context = /* a deep_causality::Context */;

let verdict = ethos
    .evaluate_action(&action, &context, &["flight_safety".to_string()])
    .expect("evaluation failed");

match verdict.outcome() {
    TeloidModal::Impermissible => { /* forbidden */ }
    TeloidModal::Obligatory    => { /* required */ }
    TeloidModal::Optional(_)   => { /* permitted with cost */ }
}

for norm_id in verdict.justification() {
    if let Some(norm) = ethos.get_norm(*norm_id) {
        println!("#{} {} -> {:?}", norm_id, norm.action_identifier(), norm.modality());
    }
}
```

A full worked example, including the `Context` setup and a CSM integration, lives at
[`examples/csm_examples/csm_effect_ethos`](../examples/csm_examples/csm_effect_ethos).

## Modalities

| Modality           | Meaning                                                                |
|--------------------|------------------------------------------------------------------------|
| `Obligatory`       | The action must be taken; omission is a violation.                     |
| `Impermissible`    | The action must not be taken; performing it is a violation.            |
| `Optional(i64)`    | The action is permitted and carries an explicit cost.                  |

## Relation to other DeepCausality crates

* `deep_causality` supplies `Context`, `ProposedAction`, `Uncertain`, and the seven generic parameters used here.
* `ultragraph` backs the `TeloidGraph`; freeze and acyclicity checks come from it.

## References

* Olson, T., Salas-Damian, R., and Forbus, K. D. *A Defeasible Deontic Calculus for Resolving Norm Conflicts.* Department of Computer Science, Northwestern University. The DDIC formalism underlying the conflict resolution rules used here:
  [docs/papers/ddic.pdf](../docs/papers/ddic.pdf)
* Effect Propagation Process paper, section 8 (Teleology):
  <https://github.com/deepcausality-rs/papers/blob/main/effect_propagation_process/epp.pdf>
* In-repo overview: [docs/ETHOS.md](../docs/ETHOS.md)

## License

This project is licensed under the [MIT license](LICENSE).