Implicit feedback
Implicit feedback is indirect evidence about user preferences or judgments inferred from observed behavior rather than explicitly stated ratings or evaluations.
Building Reliable AI Systems
Implicit feedback is evaluative information inferred from user behavior rather than directly supplied by users.
- How it is used
- Turns search clicks into embedding-training pairs and uses interaction patterns to monitor whether deployed AI responses help users.
- How deep it goes
- Identifies behavioral signals and click-position bias, but provides no debiasing algorithm or signal-collection code.
- Examples
- Illustrates search clicks, abandonment, rapid rephrasing, copy-paste, return frequency, and repeated follow-up questions.
- What you can do after
- Diagnose/debug with it by treating user behavior as a response-quality clue and checking whether position bias distorts search-click training data.
Read how Building Reliable AI Systems explains implicit feedback →
Reinforcement Learning from Human Feedback
Implicit feedback is information about user preferences inferred from users' actions rather than explicit evaluations.
- How it is used
- Uses user behavior to tune chatbot and model performance when explicit preference data does not fully capture downstream user needs.
- How deep it goes
- Defines behavioral signals and their optimization risk, but stops short of specifying a settled measurement or training procedure.
- Examples
- Uses re-rolling outputs, closing tabs, and writing angry messages as behavioral quality signals.
- What you can do after
- Evaluate/understand response quality from user actions and recognize the stability risks of optimizing those signals.
Read how Reinforcement Learning from Human Feedback explains implicit feedback →
Related concepts
- monitoringused by · Monitoring is the ongoing observation and measurement of a system’s behavior, state, or operational signals to assess its condition and detect changes or problems.
- feedbackkind of · Feedback is information about an action, system, or outcome that can inform evaluation, decisions, or subsequent behavior.
- click modelproduced by · A click model is a predictive model of the likelihood that a user clicks a particular result, advertisement, or other presented item.
- customer satisfactionmeasures · Customer satisfaction is a measure of how favorably customers or users evaluate a product, service, interaction, or outcome in relation to their needs and expectations.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.