A Feed Control Is a Preference, Not a Conversation

An interface that accepts an ordinary sentence can feel conversational. Facebook’s October “Your algorithm” announcement invites people to say what they want more or less of in Feed and Reels. Meta says those preferences will help its ranking systems adapt. That is a product description, not evidence that a person is talking to a reviewer, receiving an explanation, or opening a case about a particular post.

The distinction is practical. “Show me more local history” is a preference about future recommendations. “Why did this post appear?” asks for an explanation. “This decision restricted my post; please review it” contests a specific action. A feed may offer all three kinds of control, but they do different work and require different evidence.

A preference can be explicit without becoming a case

Meta’s October 13 announcement says Facebook is rolling out a space where users can add and manage topics for Feed and Reels, plus more ways to give feedback from a post. After a person selects “Interested” or “Not interested,” Meta says it may ask follow-up questions and allow the person to describe what they want in their own words. The company says large language models help interpret interest, context, and intent, and that recommendations can adapt within the same session. These are Meta’s claims about the feature and its ranking system; the announcement is not an independent evaluation of how well it interprets requests. Meta’s Your algorithm announcement

An explicit preference differs from an inference. A sentence is a stated request; a like, comment, or watch is behavior from which a system may infer interest. Meta says the new freeform text is combined with signals it already uses. That makes the control a direct input to a ranking process. It does not make the interface a human conversation, nor does it establish that the person can inspect the calculation behind a particular result.

The announcement describes adaptation, but it does not promise that every request will prevail over other signals, disclose a trace for each recommendation, or say that a request creates a review record. Those gaps do not prove the feature fails. They define what the public description cannot establish. If someone sees a post they hoped to avoid, that single result cannot reveal which signal shaped the ranking or whether the stated preference was ignored.

Feedback is not the same as enforcement

Product feedback concerns the experience a person wants. A report or moderation appeal concerns whether a particular piece of content or a platform decision violates a rule. A user can dislike an ordinary post without alleging that it is prohibited. A user can also believe a post violates platform policy even if the user would otherwise want to see similar content. Preferences and policy judgments are separate dimensions.

Meta’s corporate Human Rights Policy says the company gives people an option to appeal a broad range of content-moderation decisions and may restore content that was incorrectly removed or remove content that was incorrectly left up. The policy points to transparency reporting about those decisions. That is Meta’s high-level account of a remedy process, not a description of every eligible decision, the steps for a specific account, or an outcome guarantee. Meta’s Corporate Human Rights Policy

This matters for the word “feedback.” A conversational box may invite more detail than a binary button, but the detail can still be used to tune a feed rather than ask a person to reconsider a decision. Meta’s announcement calls it a way to shape the experience. It does not say that freeform feed text becomes a complaint, reaches a human reviewer, or triggers a statement of reasons. Readers should not infer those functions from the conversational style of the interface.

An appeal needs a decision to contest

The European Commission’s current summary of the Digital Services Act draws a separate line around content moderation. It says platforms must explain when content or an account is removed or suspended, and users can appeal content-moderation decisions through a platform system or an out-of-court dispute-resolution body. The Commission separately describes the feed option for very large platforms as a choice to use a non-personalized feed. One is a way to choose among recommendation approaches; the other is a route to challenge a moderation decision. The Commission’s Digital Services Act overview

An appeal starts with an identifiable action: a removal, restriction, suspension, or other moderation decision that a person can challenge under the relevant process. A recommendation preference starts with a desired experience. If a platform reports that content was removed or distribution restricted, a person can look for the notice and the available review mechanism. If the person only wants fewer recommendations on a topic, an “interested” control may be the relevant input, but it is not a record that a caseworker heard the complaint.

What a conversational control should make legible

A useful interface can accept open-ended input and still clarify what follows. A reader should be able to tell whether a message changes ranking, reports a policy concern, or requests review of an adverse decision. Those paths can sit beside one another, but they should not be disguised as one.

Four questions help keep the states distinct:

• What did the person do? State a topic preference, give ranking feedback, report content, or appeal a moderation action? A button label or confirmation screen should identify the operation.
• What happens next? Does the input alter future recommendations, open a report, or start a review? Meta’s public product post describes the first function; it does not document the latter two for freeform feed requests.
• What can be checked? A preference tool could show what topics were saved and how to edit or remove them. A moderation process should identify the decision and the review route. These are separate forms of legibility, not claims about what Facebook currently provides.
• Who is accountable? Ranking remains a platform system even when a person can steer it. An appeal requires an accountable process for reconsidering a defined decision, not merely an interface that responds in friendly language.

Natural-language controls can make personalization easier to express. That is a real kind of agency, but it is limited to the job the feature is designed and documented to do. Calling the interface a conversation should not make readers assume there is a listener, an explanation, or an appeal behind every answer from the feed. A preference can influence what comes next; accountability begins when the platform makes its decisions and remedies separately visible.

Sources and limitations

• Meta, “Your Algorithm, Your Words: New Ways to Shape What You See on Facebook” — checked October 21, 2026; supports Meta’s description of topic controls, natural-language feedback, and ranking adaptation for Feed and Reels. Limitation: company announcement, not independent testing; it does not describe a review or appeal workflow for feed preferences.
• Meta, Corporate Human Rights Policy — checked October 21, 2026; supports Meta’s statement that it offers appeals for a broad range of content-moderation decisions and reports related data. Limitation: high-level company policy; it does not enumerate every eligible decision or establish outcomes for a particular feature.
• European Commission, The Digital Services Act — checked October 21, 2026; supports the Commission’s explanation of content-moderation appeal options and non-personalized feed choice for very large platforms. Limitation: Commission overview, not a decision about Facebook’s new preference control or any individual case.