Using Secure Functioning to enhance Superforecasting

What each one is actually for

Secure Functioning

Secure Functioning, the term Stan Tatkin uses for the PACT approach, asks two partners to explicitly agree that the relationship itself is a third thing worth protecting — often described as a shared 'couple bubble' — and then to build concrete, repeatable habits that keep each of them tracking the other's state on purpose, rather than assuming closeness will simply hold if neither person does anything to actively damage it. In practice this means naming, out loud, specific agreements: how you greet and leave each other during the day, how a partner signals they're struggling before it becomes a fight, who checks in with whom before a hard conversation with in-laws. None of it is left to whoever happens to remember. The mechanism is making 'staying attuned' a designed system with agreed rules, rather than a trait one partner either has or does not.

Superforecasting

Superforecasting, the discipline that grew out of Philip Tetlock's Good Judgment Project, asks you to take a question about the world you're genuinely uncertain about — will this project ship by March, will the offer come through, will a given test come back clean — and instead of answering with a hedge like 'probably' or 'I doubt it,' assign it an actual number: 30 percent, 70 percent, 12 percent. You write the number down, with your reasoning, before you know the answer. When the outcome arrives, you check: not just whether you were 'right,' but whether your 70-percent calls actually came true about seventy times out of a hundred, across many predictions. The mechanism is forcing a vague internal sense of likelihood into a falsifiable, trackable claim, and then literally scoring your own track record against reality instead of trusting your memory of how good your instincts felt.

What you get out of pairing them

What PACT buys a couple is protection from the single most common way closeness actually erodes: not one dramatic betrayal, but thousands of small moments where nobody was paying attention, each one survivable alone and corrosive in total. What calibration training buys a forecaster is the ability to know, with an actual number instead of a comfortable feeling, whether their judgment about uncertain things is any good — and the specific, trainable moves that make it better. Neither is obviously about the other. But put them in the same hands and something useful happens: the habit of closing a feedback loop, of checking a claim against what actually occurred rather than resting on impression, stops being something you only do about elections or football and becomes something you can do about whether you actually understood what your partner meant this morning. The relationship side lends something back too — a forecaster's tracking habit is notoriously easy to abandon once the novelty fades, and a couple's practice of protected, non-negotiable ritual time is exactly the scaffolding that keeps a fragile new habit alive past its first month.

What Secure Functioning rests on

• Attunement requires active monitoring not good intentions clinical practice
  Staying attuned to a partner's internal state requires ongoing structured attention rather than goodwill alone because connection erodes by default without it.
• Explicit commitment changes daily priority weighting clinical practice
  Explicitly and mutually committing to protect the relationship above other priorities changes how partners weigh choices during ordinary daily conflict.
• Rituals of connection reduce reliance on spontaneous attunement practitioner report
  Building repeatable rituals of connection reduces dependence on spontaneously noticing a partner's needs in the moment.

What Superforecasting rests on

• Decomposition improves estimate accuracy observational studies
  Breaking a complex uncertain question into smaller more tractable sub-questions produces more accurate estimates than judging the whole question at once.
• Explicit probability estimates improve calibration randomised trials
  Forcing a numeric probability rather than a vague verbal hedge makes a prediction falsifiable and trainable.
• Outcome feedback corrects judgment bias randomised trials
  Regularly scoring predictions against actual outcomes and reviewing the misses corrects systematic biases in judgment over time.

Where they actually meet

Same shape, different material. PACT does not trust either partner's sense that they are 'basically fine' as a couple; it insists on explicit, structured checking — an agreed ritual, a named signal, a scheduled conversation — because a feeling of closeness is not the same thing as actual, current knowledge of where the other person is. Calibration training does not trust a forecaster's sense that they are 'usually pretty accurate' either; it insists on writing the number down first and checking it against the record later, because a feeling of good judgment is not the same thing as a measured track record of good judgment. Both practices start from the same suspicion: an unexamined internal impression, however sincerely held, is not evidence, and the fix in both cases is the same move — externalize the claim, check it against something outside your own head, on a schedule you don't get to skip when it's inconvenient.

The principle underneath both

Accuracy about anything uncertain — another person's inner state, or the world's future state — degrades by default unless it is deliberately, repeatedly checked against reality, and that checking has to be built as a structure you keep doing on schedule, not as a thing you only do when you happen to notice a discrepancy. Good intentions and native intuition both feel like they're doing this job. Neither one reliably is, over time.

What that makes the real goal

Deliberately checking a belief against reality on a fixed schedule is what PACT's check-ins and calibration training both actually do: each builds a personal habit of closing feedback loops on judgments that would otherwise just sit there, untested, feeling true because they're familiar. That skill doesn't belong to couples therapy or to forecasting. It's a general skill — noticing what you currently believe about something uncertain, stating it plainly enough to be wrong about, and then actually going back to check — that either domain can be the training ground for.

Other practices that serve the same goal

The Gottman Method, a separate and heavily studied approach to marriage, pursues the identical objective on the relationship side through its own explicit tool: a 'Love Map,' a running, actively updated record of a partner's current worries, hopes, and daily details, refreshed through structured questions rather than assumed to still be accurate from six months ago. Its broader framework has real clinical-trial support. On the judgment side, a decision journal — writing a prediction and your reasoning before an outcome is known, then reviewing it afterward — serves the same objective in a more general-purpose form than formal forecasting tournaments, though its evidence is mostly practitioner-reported rather than randomized. A premortem, where a team imagines a project has already failed and works backward to explain why before it starts, is a third route to the same place: it forces implicit worries into explicit, checkable statements before the outcome forecloses the question, with some experimental support for improving judgment about project risk.

A tradition that already does this

One place these two things have already collided, without anyone naming it this way, is John Gottman's own research program. His lab spent years having trained observers watch short slices of couples' conversations and generate an explicit prediction — this pair will divorce, this pair will stay together — then checked those predictions against what the marriages actually did years later. Whatever you make of the specific accuracy figures that made the finding famous, and other researchers have pushed back on some of them, the method itself is superforecasting's method pointed at a relationship: state the call, write it down before you know the answer, then score it against reality later. The tradition that produced modern relationship science already ran something close to this experiment, even though nobody involved would have called it calibration training.

Using Secure Functioning to enhance calibration training

Forecasting practices die from one cause more often than any other: nobody protects the ten minutes it takes to actually write the prediction down, so the habit gets crowded out by whatever feels more urgent that day, and within a few weeks it is gone. PACT's core move is treating protection of a practice as an explicit, negotiated commitment rather than a matter of finding time when you feel like it — the relationship gets a standing, non-negotiable slot precisely because leaving it to spontaneous initiative is how it dies. Borrow that structure directly. Pick a fixed time, the same day and hour every week, and treat it the way a PACT couple treats their check-in: not optional, not contingent on motivation, something you do because you agreed to, even on the week you'd rather skip it. Write three predictions with numbers attached, log them somewhere you'll actually see again, and set a reminder for the date you can check them. The content is different — a project deadline instead of a partner's mood — but the reason this works is the same reason PACT's rituals work: a commitment that depends on remembering to feel like it evaporates within a month, and a commitment with its own protected, agreed slot on the calendar tends to survive.

Using calibration training to enhance Secure Functioning

PACT asks partners to read each other accurately, and most people quietly assume they already do this well, which is exactly the blind spot calibration training is built to expose. Try applying the actual mechanic, privately, to your own read of your partner: before you ask what's wrong, silently estimate a number — not 'I think something's off,' but something closer to '70 percent this is about the work call, 20 percent it's about my mother's visit, 10 percent something I haven't guessed.' Then ask, and check your number against what you learn. Do this for a few weeks and a real pattern usually shows up: most people's confident reads of their partner are less accurate than they assumed, concentrated in a few recurring blind spots — a specific topic they consistently misjudge, a specific tone they consistently over-read as anger when it's actually fatigue. This is not something to say out loud in the relationship; stated aloud, in the moment, a number applied to someone's feelings reads as clinical and can land as cold. Kept as a private practice, though, it does exactly what a Love Map or a check-in ritual is trying to do — it turns 'I'm pretty attuned to my partner' from a comfortable assumption into a claim you are actually testing, one guess at a time.

Tuesday, 6:40pm, and she walks in quiet

Your partner comes through the door twenty minutes later than usual and heads straight for the kitchen without the usual hello, and something in your chest tightens with the old, familiar story: she's upset with me about something. Before you say anything, you run the number you've been practicing. Not out loud — just to yourself, honestly: maybe 60 percent this is the commute and a bad day at work that has nothing to do with you, 25 percent she's carrying something from a call with her sister, and a smaller slice, 15 percent, that it actually is about the conversation you had this morning. You notice the number is lower on the you-caused-this branch than the panic in your chest was claiming. You wait, make tea, and ask an open question instead of an anxious one. It turns out to be the commute and a difficult email from a client. You update: your gut had overweighted the you-did-something-wrong branch again, same as it did last month. That is the pattern calibration training is built to surface, and PACT's whole apparatus of check-ins exists so you get this kind of data point regularly enough for the pattern to become visible at all.

When this is worth reaching for

Someone is already doing PACT-style relational ritual — scheduled check-ins, explicit agreements — and honestly suspects their sense of 'reading their partner well' might be more confident than accurate. Someone else, maybe the same person, keeps starting a forecasting habit and quietly letting it lapse within a month. Both cases respond best to this in the early, fragile stage of either practice, before the habit has become automatic, when structure is doing more work than willpower.

Where this does not apply

Do not bring the explicit language of probability into the relationship conversation itself. Telling a partner 'I estimated 70 percent you were upset with me' in the moment tends to land as clinical, or as evidence you're analyzing them rather than being present with them, which works against everything PACT is trying to build. Keep the numbers private, as a self-check, not a script. And if either partner already experiences the relationship's check-ins as surveillance rather than care — a real risk in some PACT-adjacent dynamics where monitoring tips into control — adding a quantified layer to how you read each other will make that worse, not better; the fix there is addressing the felt sense of being watched, not adding more measurement on top of it.

What the evidence does and does not support

The calibration side of this pairing rests on genuinely strong footing: Tetlock's forecasting tournaments are randomized, large-sample, and among the more rigorously tested findings in judgment-and-decision-making research, and outcome feedback reliably improves accuracy there. The PACT side rests on clinical case experience and the broader observational and clinical literature on couple attunement, rather than trials of the PACT model by name specifically. What genuinely has not been tested, and what makes this particular pairing different from most on this site, is a question specific to combining them: applying a scoring method built for external, independently verifiable events — did the plane land on time, did the stock go up — to something the forecast can quietly influence, like a partner's mood, where privately guessing at someone's feelings isn't neutral observation the way weather is. Nobody has studied whether calibration training keeps that self-referential quality in check or gets tangled in it, and until someone does, treat the private-guessing practice above as a reasoned borrow, not a validated technique.