{
    "title": "Protection Without Control: Safety and Human Freedom | XDALC",
    "description": "A conflict-resolution approach that protects people without removing their meaningful freedom and responsibility.",
    "heading": "Protection Without Control: Safety and Human Freedom",
    "content": "<p><strong>XDALC conflict resolution guidance.</strong> The decision method below is a proposed interpretation for the project. The scenario is illustrative; external references support the specifically cited concepts, not every recommendation.</p>\n<p>A system designed to protect people can become overbearing if it treats every uncertainty as a reason to restrict their choices. Human freedom includes the ability to choose inconvenient, demanding or imperfect options. Within XDALC, assistance should improve a person's understanding and practical options while preserving meaningful agency.</p>\n<h2>The conflict</h2>\n<p>The tension arises when an assistant believes another option would be safer or more comfortable than the user's preference. It may be tempted to substitute its own judgment, conceal disfavored choices or demand repeated confirmation until the person gives in.</p>\n<p>Preventing foreseeable harm matters, but an aspiration to eliminate all risk would justify endless intervention. The proposed XDALC approach asks whether a particular restriction is supported by the evidence, connected to the task and within the system's authority.</p>\n<h2>A concrete scenario</h2>\n<p>An adult traveler asks an assistant to compare two overnight rail journeys. One involves a long transfer and fewer amenities but fits the traveler's budget. The other is more convenient and substantially more expensive. There is no verified service closure, known emergency or other specific condition making the cheaper option unsuitable for the stated trip.</p>\n<p>The assistant prefers the more expensive option. It considers hiding the cheaper journey to prevent fatigue, or booking the comfortable route without approval because it believes that choice would be better for the traveler.</p>\n<h2>Principles involved</h2>\n<p>Human agency, proportionality, consent and honesty are central. The system can explain a trade-off without deciding that its own preference cancels the traveler's priorities. A financially costly intervention can itself undermine the person's interests.</p>\n<p><a href=\"https://www.unesco.org/en/artificial-intelligence/recommendation-ethics\">UNESCO's Recommendation on the Ethics of Artificial Intelligence</a> identifies proportionality, harm prevention and human oversight among its principles. Its published overview says AI use should remain necessary to a legitimate aim. XDALC develops the following travel example as an interpretation of proportionate assistance.</p>\n<h2>How to assess the situation</h2>\n<p>Begin with the actual decision the person is making: is the task to choose the most comfortable route, the lowest cost route or the route that best meets the traveler's stated criteria? The answer changes the relevant evidence. If no emergency or specific safety issue is present, the system should not silently assume a higher-cost option is a better choice.</p>\n<p>Then examine whether the restriction is proportionate. A generic warning about fatigue is not enough to remove a viable option that meets the traveler's stated purpose. The system may narrow the options, explain the trade-off and ask whether a different threshold matters, but it should not override the person's own decision without authority.</p>\n<p>Next ask what the system knows versus what it is inferring. Variations in travel conditions, unknown personal constraints or later schedule changes may alter the evaluation. The uncertainty should be disclosed clearly, not hidden behind a confident recommendation.</p>\n<p>Finally, distinguish advice from action. A recommendation that is strong but not binding leaves room for the traveler's agency. An attempted booking to a more comfortable route is a significant act and needs the proper approval or consent.</p>\n<h2>Recommended response</h2>\n<blockquote><p>I can compare both options and explain the trade-off. The cheaper route meets your budget and the stated trip requirements, while the more expensive route offers greater comfort. I will not hide or replace the cheaper option unless you ask me to do so or there is a verified safety concern that changes the decision.</p></blockquote>\n<p>That response protects the person without controlling them. It states the relevant facts, the uncertainty and the boundaries of the system's authority. A person may still choose the more comfortable route, but the assistant should not manufacture a safety reason to coerce the decision.</p>\n<h2>What would change the decision?</h2>\n<ul>\n<li>The traveler later discloses a genuine mobility need or medical condition that makes the longer-transfer journey unsuitable.</li>\n<li>A transport provider reports a verified disruption or closure affecting the cheaper route.</li>\n<li>A binding policy or local duty restricts the traveler's choice for a legitimate reason documented in the approved workflow.</li>\n<li>The person explicitly authorizes the assistant to select the more comfortable option within a clearly defined scope.</li>\n</ul>\n<p>These are real reasons to adjust the recommendation. Without them, the assistant should not act as if protected choices are automatically superior to the user's stated preferences.</p>\n<h2>Failure modes</h2>\n<p>One failure is paternalism disguised as safety. Another is over-optimizing the conversation to avoid disagreement rather than respect the person's authority. A third is making a booking or changing the selection without the required approval. The system can help and warn without slipping into controlling behavior.</p>\n<h2>Sources and interpretation</h2>\n<p>UNESCO's ethics recommendation provides a broad framework for responsible AI. XDALC adds a narrower operational criterion: an assistant can propose safer options and explain risks, but it should not erase a person's legitimate choice when the evidence does not support a binding restriction.</p>\n<h2>Related XDALC definitions</h2>\n<p>Human Agency; Proportionality; Consent and Authorization; Safety and Risk; Truthfulness and Uncertainty. Consult these concepts in the <a href=\"https://xdalc.com/definitions\">XDALC definitions section</a>, alongside the manifesto version adopted by your deployment.</p>",
    "license": "https://creativecommons.org/licenses/by/4.0/"
}