Five items measuring the overall judgement a person makes about their life. Stable, well understood, and close to useless alone for evaluating an AI system.
| Full name | Satisfaction With Life Scale |
| What it measures | Global life satisfaction. The judgement a person makes about their life as a whole |
| Items and time | Five items, about a minute |
| Response scale | Seven points, from strongly disagree (1) to strongly agree (7) |
| Recall period | None. Life as a whole |
| Score range | 5 to 35 |
| Developed by | Diener, Emmons, Larsen and Griffin, 1985 |
| Cost and licence | Free with credit to the authors. Diener’s newer site says non-commercial use only, his older university page calls it public domain |
| Languages | More than 40 |
| Ages | Adults and adolescents. A children’s version exists |
| Best used for | A life-satisfaction module inside a wider design. Long horizons |
| Pair with | Anything that measures functioning or capability, because this measures neither |
Five statements, a seven-point agreement scale, and forty years of use. The SWLS is probably the most used single wellbeing scale in the world, and it is translated into more than forty languages. It asks a person to step back and judge their life. That is all it asks, and it does it very well. It is also the measure most likely to be misread, because the word satisfaction means one thing here and something quite different on a product dashboard.
Diener’s sites permit reproduction with credit, so here they are.
Each is answered strongly disagree (1), disagree (2), slightly disagree (3), neither agree nor disagree (4), slightly agree (5), agree (6), strongly agree (7).
Notice what kind of question these are. None asks how you feel. Each asks for a verdict, and the last one asks for a verdict on the whole thing in hindsight. That is deliberate. Diener and his colleagues wanted the cognitive, evaluative side of subjective wellbeing, separate from mood, and they built a scale that would let a person weigh their life by whatever standards they hold rather than by standards the researcher chose. The fifth item is the one that behaves a little differently from the others in many samples, which makes sense given that it asks about the past.
The scale was published in 1985 in the Journal of Personality Assessment. The authors started with 48 items, cut them to the ten that loaded on a life satisfaction factor, and then to five after removing redundancy. It was validated on two undergraduate samples and a sample of older adults, with convergence checks against interviewer ratings, other wellbeing measures and informant reports. Pavot and Diener’s 1993 review in Psychological Assessment pulled together the first decade of evidence and remains the paper people cite. Ed Diener died in 2021, and his family maintain the scales online.
Add the five items. The total runs from 5 to 35, and higher is more satisfied. No reverse scoring, no weighting, no conversion. If an item is missing, the common practice is to average the remaining four and multiply by five, though the original paper gives no rule.
Diener’s scoring sheet gives seven bands: 31 to 35 is extremely satisfied, 26 to 30 satisfied, 21 to 25 slightly satisfied, 20 is the neutral point, 15 to 19 slightly dissatisfied, 10 to 14 dissatisfied, and 5 to 9 extremely dissatisfied. His longer note on interpreting scores draws the bands slightly differently (30 to 35 very high, 25 to 29 high, 20 to 24 average, and so on down) and says that the average in economically developed nations falls in that 20 to 24 range.
So the typical adult in a country like ours scores as slightly satisfied to satisfied, mostly content with most areas of life and seeing room for improvement in some. Group means in the low to mid twenties are ordinary. A group mean in the high twenties is doing well. Below 20 as a group mean is unusual outside clinical or very disadvantaged samples and worth taking seriously.
Change is a different matter. The scale is built to be stable, and it is. A difference of a point or two between two readings of the same group is within what the scale’s own stability would produce, and there is no agreed threshold for a meaningful change the way there is for the WHO-5. If you want to see life satisfaction move, you need either a long interval or a large event.
In the original paper the internal consistency was 0.87 and the two-month test-retest reliability 0.82, with a single factor accounting for most of the variance. Pavot and Diener’s review reported internal consistencies across studies in the high 0.7s to high 0.8s and test-retest figures that fall as the interval lengthens, down to 0.54 over four years, which is what you would expect of a measure that is stable but not fixed.
The scale converges with other wellbeing measures, with informant reports and with clinical ratings, and it is distinct from measures of emotional wellbeing, which is the discriminant evidence that it is measuring judgement rather than mood. It has been used in national panels, in clinical trials and in cross-national comparisons, and the single-factor structure has mostly held up across translations, with the fifth item the usual exception.
The five-item adult scale is the standard. There is a children’s version, the SWLS-C, validated for ages around nine to fourteen, and a temporal version that asks the five questions about the past, present and future separately. Translations exist in more than forty languages and many of them have their own validation papers.
On permission, Diener’s older University of Illinois page says the scale is in the public domain and may be used by researchers and practitioners without permission or charge as long as the authors are credited. The newer eddiener.com site says the scales are copyrighted and free for non-commercial use with credit. For anything on a non-commercial site like this one, either reading allows use with a citation. For a commercial product, ask.
It is a minute long, it is free, it is translated, and it is understood by everyone in the field, so a result on it needs no explanation. It is stable, which makes it a good anchor and a poor sensor. It measures something that matters in its own right, how people judge their lives, and it lets them do the judging. And its psychometrics have been examined more thoroughly than almost any other brief scale.
It measures one component of wellbeing: the judgement, not the feeling, and not the functioning. Nothing in it concerns relationships, meaning, capability or growth except as those happen to inform a person’s overall verdict. As a stand-alone outcome it is narrow.
Life satisfaction can stay high, or rise, while other things are going wrong. People adapt. Convenience is pleasant. And the judgement the scale asks for is made from inside the situation being judged. This is the clearest case in the whole set of a measure that will look fine while capability, autonomy or relationships erode, which is exactly the pattern anyone evaluating an AI system should be worried about. A system that makes life easier will not lower anyone’s life satisfaction in the short run and may raise it, and that tells you almost nothing about what the system is doing to the person over time.
Then there is the confusion of words. Satisfaction with your life is not satisfaction with a product, a service or an AI system. The second is a customer metric, measured with different questions, and it tells you about a pleasant immediate experience. A dashboard that reports satisfaction with a system as though it were evidence about wellbeing has swapped one construct for another, and it happens all the time.
It is sensitive to mood on the day, within limits, and to social desirability if people think their answers are seen by anyone with authority over them. And like every self-report scale, it cannot attribute a change to a cause.
I would include it when I wanted a brief, well-understood reading of how people judge their lives, as a long-horizon module in a wider design. The Flourishing Measure already contains a life satisfaction item, so if that is in the battery the SWLS is adding depth rather than coverage. On its own, for evaluating an AI system, it is close to useless, because it measures the one thing most likely to look fine. Pair it with a measure of functioning or capability, and with something specific to the setting.
The scale, the scoring sheet and the note on interpreting scores are at eddiener.com/scales/7 and the SWLS page, eddiener.com/satisfaction-with-life-scale-swls. The older University of Illinois page, with the public domain statement, is at labs.psychology.illinois.edu/~ediener/SWLS.html.
The original paper is Diener, E., Emmons, R. A., Larsen, R. J. and Griffin, S. (1985), The Satisfaction With Life Scale, Journal of Personality Assessment, 49(1), 71–75, doi.org/10.1207/s15327752jpa4901_13. The review is Pavot, W. and Diener, E. (1993), Review of the Satisfaction With Life Scale, Psychological Assessment, 5(2), 164–172.