The Moment Is Not the Answer

This series examines just-in-time adaptive interventions, or JITAIs: intervention designs that use changing information about a person and their surroundings to decide whether and how to offer support. The question is not merely whether a digital tool can deliver a message quickly. It is whether the message, exercise, reminder, or other option fits the person’s situation well enough to help. That distinction gives readers a way to understand what JITAIs are, what their evidence means, and why a timely intervention can still be the wrong intervention.

The problem with a message that is always ready to send

Many digital interventions begin with a sensible intention: make useful support available outside a clinic, classroom, or scheduled appointment. A phone can remind someone to move, invite them to notice a difficult emotion, suggest a coping strategy, or encourage a health behavior. But availability is not the same as suitability. A person’s need, attention, surroundings, and ability to act can change from one hour to the next.

Consider a simple contrast. Someone trying to establish a health routine may benefit from a reminder before the usual time for that behavior. Yet the same reminder can arrive while they are driving, rushing into a meeting, caring for someone else, or already carrying out the behavior. A fixed message treats these moments as equivalent because the clock says they are. A JITAI treats the clock as only one possible part of a decision. It asks what is happening now, whether that situation signals a meaningful opportunity or risk, and whether the person can use the particular support being considered.

This matters especially for mental health and behavior change. Stress can increase the near-term risk of a lapse for someone trying to stop smoking, but stress can also make it harder to read and process a phone prompt. A person who has just completed a self-monitoring task may be in a useful position to receive feedback or reinforcement, while the same feedback delivered at a random time may be easy to ignore. A location, a movement pattern, a self-report, or a recent behavior may provide a clue about the moment—but none of those clues is the whole person or the whole cause of what happens next.

The fixed-message model therefore has two weaknesses. First, it can deliver support when the relevant problem is not present. That creates unnecessary effort and can make repeated prompts feel like noise. Second, it can deliver the right general kind of support at a moment when the person is unable or unwilling to engage with it. A message can be accurate and still fail because its timing, format, or demands do not fit the situation.

A JITAI is designed to address this mismatch. Its subject is not the phone, sensor, or algorithm by itself. Its subject is the intervention design that governs repeated decisions in a changing natural environment. The technology is the means by which information is gathered and an option is delivered; the important question is the relationship between changing conditions and the intervention decision.

What makes an intervention just-in-time?

The term “adaptive” can make any digital intervention sound more sophisticated than it is, so the boundary needs to be precise. A JITAI uses rapidly changing information about a person’s internal state or context to decide whether and how to intervene. “Dynamic” may refer to feelings, symptoms, behavior, location, the presence of other people, or another condition that can change during the intervention. Demographics, personality, or baseline severity may help tailor a program, but using stable information alone does not make an intervention a JITAI.

The adaptation is also protocolized. In practice, that means the intervention follows a specified decision rule rather than improvising. The rule links three elements: a tailoring variable, one or more intervention options, and a threshold or condition that determines which option is selected. A decision point is when that determination is made. It does not necessarily end with a prompt. “Do not deliver this option under these conditions” can be the intended decision.

An illustrative oral-self-care rule makes the logic visible. Twice each day, before a person’s prespecified brushing window, an electric toothbrush can provide information about whether brushing occurred in the previous 12 hours. If it did not, the mobile intervention delivers a prompt encouraging brushing; if it did, the prompt is withheld. The important feature is not the toothbrush or the notification. It is the rule’s connection between a changing behavior, a decision point, and two possible actions.

That rule also shows what a JITAI is not. It is not simply an app that sends support at a convenient time, nor is it necessarily an artificial-intelligence system that invents a new message for each user. A JITAI can be a stand-alone component or part of a larger program that includes education, coaching, medication, counseling, or on-demand content. Its defining feature is the rapid, evidence-based adaptation of delivery within the intervention protocol.

A related concept, the ecological momentary intervention or EMI, also offers support in daily life and may use ecological momentary assessment. But an EMI does not necessarily adapt the intervention to the person’s momentary state. It may deliver support at repeated or fixed times, using a simple trigger. The concepts overlap in actual studies, and published descriptions do not always make the decision rule clear. That is why the label should never do the reasoning for us. To know whether an intervention is genuinely adaptive, we need to see what information drives the decision and what happens under different conditions.

There is another distinction that prevents confusion. In ordinary language, “adapting an intervention” can mean changing an existing program for a new population or setting—for example, altering examples or activities to make a program culturally relevant. In a JITAI, adaptation usually refers to deciding whether and how to deliver an option as part of an already specified protocol. One is a development activity performed by intervention designers; the other is an intervention activity performed during delivery.

flowchart LR A["Person and context change"] --> B["Fixed message or schedule"] A --> C["JITAI decision point"] C --> D{"Does the tailoring signal meet the rule?"} D -->|Yes| E["Offer a selected intervention option"] D -->|No| F["Withhold this option"] E --> G{"Can the person use this option now?"} G -->|Yes| H["Engagement with the option"] G -->|No| I["Low receptivity, burden, or possible harm"] F --> J["No delivery is a deliberate outcome"] H --> K["Proximal response"] I --> K K --> L["Longer-term outcome"] B --> M["Timing alone does not establish fit"]

The diagram captures the central caution. A JITAI adds a decision between changing circumstances and intervention delivery, but that decision is not automatically wise. It can be based on a weak signal, a threshold chosen without enough evidence, or a state that indicates need but not receptivity. The rule may correctly identify a vulnerable moment while choosing an option that is too demanding, too generic, or impossible to use in context.

The real question is whether the support fits

The phrase “just-in-time” is attractive because it promises a remedy for delay. Yet the deeper goal is not to act at the earliest possible second. It is to match an intervention option to a state in which it can produce a useful change. That requires at least three questions.

First, is there a state of vulnerability or opportunity? Vulnerability means a condition associated with near-term risk of an adverse outcome; opportunity means a condition in which support may promote a desirable outcome. The same condition can be both. Not drinking enough fluid in the previous hour may signal risk for falling short of a health goal, while also creating an opportunity to strengthen a drinking habit through a reminder.

Second, can the selected option affect the mechanism that matters? A prompt might address forgetfulness, while feedback might support self-reflection, and a brief skill exercise might help someone manage a difficult emotion. If the option has no plausible connection to the proximal change the intervention needs, precise timing cannot rescue it. A long-term outcome such as improved mental health or sustained behavior usually depends on shorter-term mechanisms that can actually change at a decision point.

Third, is the person receptive to this option now? Receptivity is not a permanent quality of the individual. Someone may be able to read a short reminder but not complete a multi-step exercise, or may be open to a coaching conversation but not to an automated prompt. Being stressed, driving, in a meeting, preoccupied, or surrounded by other people can make a digital option difficult to use. The moment of greatest vulnerability may therefore be the moment of lowest receptivity.

These questions also explain why JITAIs have a resource-efficiency rationale. A rule can avoid asking people to invest attention when the target state is absent or when the option is unlikely to help. Fewer unnecessary prompts may reduce burden, mental fatigue, and habituation. But restraint is not automatically efficiency. A rule that withholds too often can miss an opportunity, and a system that saves messages while failing to change the intended outcome is not necessarily resource-efficient in any meaningful sense.

This series will keep that tension in view. The evidence includes promising findings, but it also includes mixed intervention definitions, incomplete descriptions of decision rules, adherence problems, and different outcome types. A claim that support can be delivered in real time is not, by itself, evidence that every implementation is genuinely adaptive or that an immediate response will become a durable clinical or behavioral outcome.

A responsible claim about JITAIs must therefore keep several levels separate. There is the promise of delivering support in daily life. There is the mechanism proposed by a decision rule. There is the observed response to an intervention option. And there is the longer-term outcome that matters to a person’s health or functioning. Evidence at one level does not automatically establish success at the next.

The practical judgment is consequently more demanding than “Does the technology work?” We need to ask: What changing state does the intervention detect? What option does it make available? Why should that option change the next meaningful outcome? Can the person use it in that moment? What burden or harm might delivery create? And does the study design actually test the claim being made?

A JITAI earns the word “just-right” only when those questions line up well enough for the intended purpose. Sometimes the answer will be a prompt. Sometimes it will be a different option, a slower human intervention, or no digital delivery at all. The value of adaptation lies not in making every moment actionable, but in making the intervention decision more responsive to the conditions that determine whether support can help.

References