Used in non-essential contexts it constitutes addictive design

What it is

The same variable-reward machinery is a feature when it carries the product's core value and addictive design when it manufactures dwell time. The dividing criterion is necessity: does the reward loop serve the user's own goal — what the user came to accomplish — rather than the behavior the product wants to occur.

Why it happens

Necessity lives in the relation between the reward and the user's goal. In messaging, when a reply arrives is genuinely unpredictable, and the message itself is what the user wants: variance is a property of the world and the loop carries core value. Randomized feed recommendations, check-in lotteries, and random drops are the opposite: variance is injected, and the reward (one more item, points, a drop) answers no goal the user brought — it serves only "one more time." The signature of a loop running outside necessity is a measurement asymmetry: the back half of the loop (opens, dwell, frequency) is fully instrumented and targeted, while the front half (what the user got) has no metric at all — nobody checks what the lottery delivered, because the answer would be embarrassing.

Where it stops holding

Necessity is relational, not a property of product categories: a news feed is essential for a professional analyst and unnecessary for a user trying to cut back; the same interface faces both, so the criterion can only land on configurability and defaults, not on blanket verdicts. Products with genuine core value can still carry non-essential loops (a to-do app with a lottery bolted on), so judge loops one by one. In regulation, "addictive design" has its own legal elements; this is a design-ethics criterion, and the two should not impersonate each other.

Applying it

List every reward loop in the product and tag three things per loop: what the reward is, where the variance comes from, and whose goal the loop serves. For non-essential loops the fix is not to strip all feedback but to remove injected randomness — fixed intervals, deterministic content — and flip defaults from on to off so those who want it opt in. Verification: after de-randomizing, track the composition of retention — remaining usage should be attributable to functional value (task completion, goal attainment) rather than open frequency; if frequency falls while completion holds, what was removed was the addictive component.

Related

• Same group: P3.01.1 Uncertain rewards sustain repeated behavior best · P3.01.2 The mechanism acts directly on reinforcement-learning circuitry
• Adjacent: P2.08 The line between persuasion and manipulation · P3.07.4 The decoupling criterion for addictiveness
• Search terms: addictive design · persuasive technology ethics · variable reward