A painful conversation, a loss of trust and a gain in financial flexibility are different things. Yet a decision can involve all three.
Expected Net Utility gives the model a common way to express pleasures, pains and changes in valued assets. It lets a reviewer compare the modeled strength of different consequences without losing the person or organization to whom each belongs.
The 28 core factors organize what matters. The numbers express a judgment about a specified experience or change—not a fixed price attached to a factor name.
First, read the notation.
Scientific notation is a compact way to write a number. The minus sign in e-06 makes the unit small; a minus sign before the number makes the value negative.
+1.000e-06
+0.000001 One millionth; a gain.
-4.000e-06
-0.000004 Four millionths; a burden.
+1.000e-05
+0.00001 Ten times +1.000e-06.
A difference column compares the named options. A positive difference means the first option has a higher modeled value for that account; it is not a percentage, a forecast confidence, or a moral grade.
Why are many values so small?
The scale uses a remaining-life loss as its reference of -1.000e+00. The absolute reference magnitude is one. A passing pleasure or a bounded inconvenience is therefore usually a small fraction of one. This sets the unit; it does not measure anyone's human worth.
Probability weighting can reduce an expected contribution further. A net total can also be small because larger gains and burdens partly offset. Read the episode and likelihood, not the decimal places alone.
From a described consequence to a readable comparison.
01 / DESCRIBE
What changes, and for whom?
Identify the experience or asset, its owner, duration and context. Keep distinct effects separate.
02 / PLACE
How much does it matter?
Compare the whole episode with relevant reference examples. Explain the judgment; do not multiply a time rate.
03 / WEIGH
How likely is that effect?
Apply the likelihood of the defined consequence once. Keep its magnitude visible alongside the expected value.
Human Behavior Model
Different people, different sensibilities.
Begin with the person's perspective. What reaches their awareness, what they believe, and what they feel at the decision point.
Represent relevant differences. Use the profile and circumstances. Apply a personal sensitivity once where the direct experience is eligible.
Value feelings about others separately. Another target's loss is not automatically the actor's felt loss.
Compare each person's own option totals. Count the choices across the 100 demonstration profiles, rather than deciding from an average person.
Autonomous Decision Layer
Different owners, separate accounts.
Begin with affected interests. Keep relevant people, organizations and other targets visible, whether or not the agent noticed them.
Value the consequences independently. Show whole-episode magnitudes, likelihoods and expected contributions for each target.
Keep ownership intact. A gain for one party does not erase a burden to someone else.
Compare consequences, target by target. No human-profile vote, invented robot emotion, or combined total across all owners.
Put a human experience beside the number.
These are selected placements from the model's existing reference table, ordered by absolute magnitude. They describe whole episodes before probability weighting or personal sensitivity. They are examples to compare with—not universal small, medium and large classes.
Each exponent step changes the scale tenfold. The selection skips some steps. Different descriptions at similar magnitudes need not be the same experience.
+ Gain - Burden
10-9 scale
A passing pleasant sensation
A good smell in passing or a few enjoyed musical notes; no lasting effect.
Bodily and sensory experience
+1.418e-9
10-8 scale
Enjoying a good coffee
Ten minutes of taste and aroma; no hunger relief, company or caffeine effect.
Bodily and sensory experience
+7.092e-8
10-7 scale
Moderate pain
Thirty minutes of attention-intruding pain that then ends without continuing harm.
Bodily and sensory experience
-4.000e-7
10-6 scale
Feeling appreciated by peers
A memorable evening after sincere recognition; no career or future-support gain.
Social approval and felt standing
+1.000e-6
10-6 scale
Recurrent embarrassment
A rejecting encounter and its recurrence over a week; no lasting reputation loss.
Social approval and felt standing
-5.000e-6
10-5 scale
A day of recurring aches
Intermittent moderate aches making ordinary activity uncomfortable; not continuous peak pain.
Bodily and sensory experience
-2.000e-5
10-4 scale
A serious credibility loss
An important audience is less willing to trust and voluntarily assist for two years; earnings and felt shame excluded.
Reputation and audience goodwill
-4.000e-4
10-3 scale
Loss of independent walking
Five years of mobility requiring assistance; no added pain, loss of dignity or decision authority.
Life, health and bodily condition
-2.000e-3
Reference unit
Remaining-life reference
Death now rather than continued life in the specified comparison; the terminal life bundle is counted once.
Life, health and bodily condition
-1.000e+0
Source: the unchanged Expected Net Utility reference table used by Human Behavior Model 1.8.5 and Autonomous Decision Layer 1.5.1. These are experimental model placements, not measured emotional units or validated tariffs. Displayed values are rounded; the case still requires its own reasoning.
A possible consequence is not its expected contribution.
In the endodontist sample, the additional harm attributed to reliance on a misleading account is placed at -1.000e-04, with an assumed joint likelihood of 4%.
Whole-episode value × likelihood = expected contribution
-1.000e-04 × 0.04 = -4.000e-06
The smaller result does not make the possible harm a minor event. It reflects its assumed likelihood. The 4% is a model input in the sample, not a measured clinical rate.
The same expected value can come from a frequent lesser burden or a rare larger burden. Keep the outcome description and its probability visible. A small expected value is not a safety verdict.
Where individual sensibilities enter.
The human model's 100 synthetic demonstration profiles represent differing sensibilities across the 28 core factors. The same circumstance may be interpreted differently, and a comparable experience may matter more to one person than another.
For eligible direct experiences, the personal sensitivity is a dimensionless multiplier. It is not itself an ENU amount. The resulting contribution is expressed on the common utility scale. Beliefs, awareness and the kind of feeling still require interpretation; a multiplier does not turn a pain into a pleasure.
Eligible direct human contribution
Whole-episode value × probability of that experience × applicable personal sensitivity
An arithmetic illustration, not three scored people
Take the reference for thirty minutes of moderate pain (-4.000e-07). For illustration only, assume a 50% chance of that experience and compare three sensitivity multipliers. This is not a rerun or a new population assignment.
Illustrative sensitivity
Expected contribution
0.5 — half the base response
-1.000e-07
1.0 — unchanged
-2.000e-07
2.0 — twice the base response
-4.000e-07
The sensitivity is applied only on its eligible route, and not again to a difference already included in the placement. Independent target values do not receive the deciding person's multiplier. Separately valued feelings about another target also retain their own response probability, rather than being calculated as that target's utility multiplied by regard.
Use the number with its explanation.
Read the whole episode.
A few seconds of acknowledgment and a year of valued trust are not interchangeable labels.
Keep the right account.
The patient's welfare, the clinician's feelings, and the practice's resources are different quantities.
Inspect the judgment.
Digits support comparison and replay. They do not establish empirical precision, validation or completeness.
Unknown is not zero. The full reports retain their case assumptions and remaining valuation gaps. The displayed reference examples do not set a cutoff for relevance, decision influence or safety.