Negative feedback must state its scope: this item or this class

What it is

After “not interested,” the system may pull only this row, or it may pause the whole class, author, and topic. If people do not know the scope, they either see the class again and judge the button dead, or find the class gone and judge the system overeager. Negative feedback scope means a reject act must declare — and honour — the unit it hits: this item, this author, this class, this source.

Item-level delete versus global reset on the profile page is grain in settings. Here it is which unit that in-feed reject hits.

Why it happens

Reject is a verb with no object by default. If the model picks the object, too small and people see no effect, too large and it collaterally damages. When scope is opaque, people cannot form a forecast of “what happens if I press again,” and use of the channel collapses to two poles: never press, or mash in anger. The dirtier semantics of the latter are a later layer.

Declaring scope turns the object into a choice. The choice also constrains learning: pick “this item” and do not update class-level features; pick “this class” before writing class-level. Scope aligned with the learning unit is what makes the negative channel rebuttable and undoable.

Studying it

The same reject, three scopes: item only, author, top-level class, stated at the act or hidden. Measure: whether later list change matches expectation, collateral damage, whether people dare use it again. Independent variables: whether scope is choosable at the act, size of the default scope. Dependent variables: match to expectation, number of collaterally hit items, reuse.

Do not take copy that says “show less like this” as scope already being clear. Watch whether the learning unit actually follows the copy. Copy says item, weights hit the class — the scope is lying.

Where it stops holding

Safety rejects (hate, porn) often must widen to class or source; send them through a separate safety channel and say so, rather than mixing them into taste “not interested.” When items have no stable class, the “this class” option cannot be offered; fall back to author or source. This entry chooses the object. It does not treat what must be visible immediately after the press, and it does not treat an over-response that wipes a class from one reject — that is a default scope chosen too large.

Applying it

• Expand “not interested” into a choosable scope: this item / this author / this class. Default small (this item); class-level wants a second confirm.
• The learning unit follows the choice, and the next screen echoes it in one sentence: “we will not recommend this title” or “we will show less of this class.”
• Check: after “this item,” other items in the class should remain; after “this class,” the class should drop clearly. If the two results swap, the scope is lying.

Related

• Same group: L6.13.1 With only positive signals the system cannot tell "not interested" from "never saw it" · L6.13.3 Negative feedback must produce an immediately visible change, or users will judge it useless and stop · L6.13.4 Users treat negative feedback as emotional expression; its semantics are imprecise · L6.13.5 Over-responding to a single negative can wipe out a whole class of content
• Nearby: L6.10 Turning Personalization Off and Resetting It · L6.07 Presenting Recommendation Reasons · L6.09 Feedback Loops and Preference Entrenchment
• Search terms: negative feedback scope · not this versus not this class · reject granularity