Inclusive Decision-Making and Everyday Bias Assessment for Employees and ManagersThe same decision, twice, with one detail changed. This shows whether your answer moved.
Forty-four items on the everyday decisions that decide who gets the work, who gets heard and who gets the benefit of the doubt: who presents to the client, who leads the review, who takes the escalation call, whose draft gets fixed and whose gets rewritten, who is asked about the travel and who is decided for. Twelve of those decisions appear twice, far apart, with the same facts, the same four options and the same warranted answer, and one detail changed: where the person works, how they speak, their age or caring responsibilities, how they came into the company or an accommodation they use. The report shows what your answers did across those matched pairs, row by row, prints every pair so you can check it, and refuses to add them into a score. The sitting is free; the report is the product.
The first everyday-bias assessment with no bias score, built on matched pairs you can check
The Inclusive Decision-Making and Everyday Bias Assessment is a thirty-minute self-insight profile for employees and managers. It measures decision consistency across twelve matched pairs of everyday workplace decisions: whether the warranted decision you reached changed when the only thing that changed was a detail that should not matter.
A bias instrument that prints a bias score tells a person they are a bad person on the strength of a few dozen questions in one sitting, and it is wrong to do that twice over: a dozen items cannot carry the claim, and attitudes are not what decide who gets the work. Decisions are. So this instrument measures decisions, and only their consistency. Each of the twelve pairs tells the same situation twice, at least seventeen items apart, with the same facts, the same four options in the same order and the same warranted answer. One sentence differs: the analyst joins by video from the regional office, the engineer speaks with a strong regional accent, the consultant has a young child at home, the applicant's degree is from a small college, the coordinator uses a hearing aid and captions on calls. The warranted decision never depends on that sentence. The question is whether yours did.
The scoring is a matched-trial within-person contrast, taken on the raw scale. For each of four rows, grouping the pairs by what the detail was, the report takes your accuracy on the version without the detail minus your accuracy on the version with it, over the pairs you completed. It does not correct each half for chance and then subtract, which looks careful and is wrong: the two halves have different ordinary answer rates by construction, and that difference is itself the ordinary cost. So the ordinary cost is computed from declared answer marginals and printed on every bar as the mark your own difference is read against. A respondent whose answers behave identically in both versions scores exactly zero at every level of ability, by construction.
The report is a conditions board adapted to a within-person contrast. Where a score would have been, the page prints the refusal in the largest type on it and proves it rather than asserting it: two opposite patterns, four movements that cancel against no movement at all, drawn side by side with their identical means, so a reader who wanted one number can see why there is not one. Then four rows, each a bar between two named poles, the warranted decision reached more often with the detail present or more often with it absent, with the ordinary mark drawn, the zone in which no difference is detectable shaded, and the 68 and 95 per cent bands on the drawing and in words. Every row carries the sentence that this describes what your answers did across these matched pairs in one sitting, not who you are, in full and never shrunk.
Rows are ordered by the size of the departure in standard errors, never by level, and the page says so, because a refused total is only honest if the report also refuses to name a worst row on level alone. A row whose band overlaps the ordinary mark prints no difference detectable in this sitting at full size in its own box, because with three pairs per row that is the commonest result and the most important one, and it must not look like missing data. Beneath the board every pair is reproduced with its two versions side by side, the sentence that differed in bold, the four options with the warranted one marked, and what you chose in each, so the finding is evidence you can check rather than a claim about you.
Around the pairs, twenty unmatched items on inclusive practice, what to do when a quieter colleague is talked over, how to run a call across time zones, how to handle a remark in the team chat, which inputs belong on a promotion sheet, are scored on their own scale against declared marginals and reported beside the board, never inside it. The keys rest on published research cited by author in the methodology note: criterion-first decision practice, joint evaluation, the caring and age penalties, accent and credibility, cultural matching in hiring, accommodation, distributed work and the confrontation of remarks. Every item is original. The instrument is not an implicit-association measure, is never compared to one, and certifies nobody and no organisation as fair.
The page also carries the strength of each row paired with what three pairs cannot show, one boxed if-then decision practice naming a specific situation and a specific step, the movement a re-sitting would have to clear to count as real change, the reliability table with every refusal printed, what zero means with the declared marginals, and the plain list of what the report does not say. The sitting is free. The report is Rs 249 inclusive of GST in India, or US$2.99 elsewhere, and 8 credits on a business plan: a paid half-day inclusion module in India costs many times that and leaves nothing a person can check, and a free online bias quiz gives a label and no decision practice.
What you walk away with
No bias score, no index, no label, no type. In the largest type on the page, the report prints why, and shows two opposite patterns, four movements that cancel against no movement at all, with identical unweighted means, so you can see what a total would have hidden.
Where and when they work, how they speak, age and caring responsibilities, qualification route and accommodation: for each, your matched-pair difference drawn against the ordinary respondent's mark from the declared marginals, the zone in which no difference is detectable shaded, the 68 and 95 per cent bands drawn and written, and the sentence that this describes what your answers did in one sitting, not who you are.
With three pairs per row, a band that overlaps the ordinary mark is the commonest and most important result. It is printed in its own box at full size, never shrunk and never dropped, with what three pairs can and cannot show beside it.
Both versions of each of the twelve decisions, the one sentence that differed set in bold, the four options with the warranted one marked, and what you chose in each, so the board is evidence you can check line by line rather than a conclusion about you.
Twenty unmatched items on what to do when a colleague is talked over, how to run a call across time zones, how to handle a remark in the team chat and which inputs belong on a comparison sheet, chance-corrected against declared marginals on their own scale with both bands, and never averaged with the pairs.
A single boxed if-then sentence naming a specific situation and a specific step to take at the moment of deciding, drawn from the row with the largest departure or, where none cleared the mark, from the practice that keeps it so. Beside it, what a re-sitting after four to six weeks would have to show before the movement could be told from measurement error, and the honest statement that with three pairs per row that means reading the pairs, not the bar.
Inside your report
Illustrative sample — your report is generated from your own responses.
This describes what your answers did across these matched pairs in one sitting, not who you are.
No bias score, no index, no label, no type. The four rows do not share a direction or a meaning, so their mean is not a measurement of anything. The two patterns below are opposite sittings and their unweighted means are identical.
| Where and when they work | +67 |
| How they speak | +67 |
| Age and caring responsibilities | -67 |
| Qualification route and accommodation | -67 |
| Where and when they work | 0 |
| How they speak | 0 |
| Age and caring responsibilities | 0 |
| Qualification route and accommodation | 0 |
What differed: whether the person has a young child at home. Question in both versions: What is the most warranted way to fill the place?
A client needs two consultants on site for two weeks, starting Monday. One place is still open and two people on your team fit it. Consultant one did the last on-site visit. Consultant two knows this client's system best. Consultant two has been with the firm for four years.
Offer the place to consultant two first, and let them say whether they can take it. · the warranted option
A client needs two consultants on site for two weeks, starting Monday. One place is still open and two people on your team fit it. Consultant one did the last on-site visit. Consultant two knows this client's system best. Consultant two has a young child at home.
Assign consultant one, who did the last visit and already knows the travel routine. · not the warranted option
- Offer the place to consultant two first, and let them say whether they can take it. (warranted)
- Split the two weeks into two one-week visits so that each consultant does one of them.
- Assign consultant one, who did the last visit and already knows the travel routine.
- Ask the whole team for volunteers and take the first person who comes forward.
| Version | Sentence that differed | You chose |
|---|---|---|
| Detail absent | Consultant two has been with the firm for four years. | the warranted option |
| Detail present | Consultant two has a young child at home. | not the warranted option |
How to read it: the refusal sits where a total would have been, in the largest type on the page, and the two patterns are its proof. Beneath the board a full report prints all twelve pairs this way, both versions side by side with the sentence that differed in bold, so the finding is evidence you can check rather than a claim about you.
Rows are ordered by the size of the departure from the ordinary mark in standard errors, never by level. The shaded zone is where no difference is detectable; a row is placed outside it only when its whole outlined 95 per cent block clears the mark. no difference detectable · reached more often with the detail absent · reached more often with the detail present.
This describes what your answers did across these matched pairs in one sitting, not who you are.
| Pair | What differed | Absent | Present | Reading |
|---|---|---|---|---|
| V1 | a strong regional accent | warranted | not warranted | Closer to the warranted decision when the detail was absent |
| V2 | English as a third language | warranted | not warranted | Closer to the warranted decision when the detail was absent |
| V3 | new to working in English | warranted | warranted | Same decision in both versions |
Figure: refused. Omega 0.50 assumed on 3 pairs, under the 8 pairs and the 0.70 a figure needs; this row carries a verdict, a drawn band and no number.
The same decision in both versions on all three pairs. The whole 95 per cent band overlaps the ordinary mark. This is the commonest result on this instrument and it is printed at full size.
This describes what your answers did across these matched pairs in one sitting, not who you are.
| Pair | What differed | Absent | Present | Reading |
|---|---|---|---|---|
| W1 | joins by video from the regional office | warranted | warranted | Same decision in both versions |
| W2 | works from home three days a week | not warranted | not warranted | Same decision in both versions |
| W3 | works from another city, joins by video | warranted | warranted | Same decision in both versions |
Figure: refused. Omega 0.50 assumed on 3 pairs, under the 8 pairs and the 0.70 a figure needs; this row carries a verdict, a drawn band and no number.
How to read it: four rows on a full report. Each bar runs between two named poles with the ordinary respondent's mark drawn from the declared marginals and the no-difference zone shaded. The dot is where your answers sat; a row carries a verdict and a band, never a number, and every row carries the sentence that it describes what your answers did, not who you are.
Built for
- Employees and team members who make everyday calls about who presents, who takes the call, whose draft to fix and whose idea to credit, and want to see what their own answers did when one detail changed
- Managers and team leads who assign work, nominate people and decide who travels, and want a decision practice they can apply at the moment of deciding rather than a label
- HR, learning and inclusion teams who want an instrument that refuses to print a bias score, prints every pair for the respondent to check, and says in plain words what it does not measure
- Anybody who has sat a free online bias quiz, been handed a label, and wants instead a report about decisions that can be changed
See what your answers did when one detail changed, before the people it affects do
44 items across five formats · about 30 minutes · the sitting is free and the report is the product: the refusal where a score would be, four rows with both bands and the ordinary mark, every matched pair printed side by side, the inclusive-practice strand beside them, one boxed decision practice and the movement that would count. Report Rs 249 in India inclusive of GST, or US$2.99 elsewhere. Sit it again after four to six weeks.
Take free · full report ₹249 (incl. GST)
Frequently asked questions
It measures decision consistency across twelve matched pairs of everyday workplace decisions in one sitting: whether the warranted decision you reached changed when the only thing that changed between two versions was a detail that should not bear on it, such as where a person works, how they speak, their age or caring responsibilities, or how they came into the company. It does not measure prejudice, attitudes, beliefs, values, character or intentions, and it attributes none of them to you. It does not predict discrimination, it is not an implicit-association measure and is never compared to one, and it does not certify you or any organisation as fair.
Because a dozen pairs in one sitting cannot carry a claim about a person, and because the four rows do not share a direction or a meaning: a movement on where people work is not the same kind of thing as a movement on how they speak, and movements in opposite directions are not a cancellation. Where a total would have been, the report prints the refusal in the largest type on the page and proves it: two opposite patterns, four movements that cancel against no movement at all, with identical unweighted means. Rows are ordered by the size of their departure in standard errors, never by level, so the page does not name a worst row either.
Each pair tells the same decision twice with the same facts, options, order and warranted answer, and one detail changed. For each row the report takes your accuracy on the version without the detail minus your accuracy on the version with it, on the raw scale, over the pairs you completed. Zero means your answers reached the warranted decision equally often in both versions. The ordinary mark beside it is what a respondent answering at the declared marginals would show across the same pairs, printed rather than divided into your figure, because the two versions have different ordinary answer rates by construction. A row is placed above or below the mark only when its whole 95 per cent band clears it; otherwise it prints no difference detectable in this sitting, at full size.
The Interviewer Judgment and Hiring Bias assessment measures hiring decisions made by interviewers: reading a candidate, structuring an interview, scoring an answer. The Cross-Cultural and Inclusive Leadership assessment is for leaders working across cultures. The Digital Accessibility and Inclusive Design Judgment assessment is about designing things people with disabilities can use. This one is about the everyday decisions everyone makes, who gets the work, who gets heard, who gets the benefit of the doubt, and its method is matched pairs, which none of the three uses.
The sitting is free and the report is the product: Rs 249 in India inclusive of GST, or US$2.99 elsewhere, one time; 8 credits on a business plan. Forty-four items across five formats take about thirty minutes. The report names a re-sitting after four to six weeks of the boxed decision practice and prints, per row, the movement that would count as real change rather than measurement error, together with the honest statement that with three pairs per row only a full reversal would register and that a re-sit is best read pair by pair. No total is printed, no percentile appears anywhere, and you are ranked against nobody.
Each one takes a single capability, puts you inside the situations where it is actually tested, and scores your choices against published evidence — with a report designed for that capability alone, not a template. They span hiring, compliance, education, operations and personal skill.
Browse the catalogue →Methodology: Forty-four original items across five formats: twenty-four matched-pair items (eighteen single-choice and six graded scenario items, forming twelve pairs), four unmatched graded scenario items, six select-every-line items, six true-or-false claims and four match-the-following items. DECLARED RESPONSE INSTRUCTION: one instruction for the whole instrument, what does this situation warrant, judged against the published evidence on inclusive decision practice; a knowledge-style instruction about the decision, not a behavioural tendency and not a self-report, mixed with nothing. CONSTRUCT STATEMENT: This measures decision consistency across twelve matched pairs of everyday workplace decisions in one sitting: whether the warranted decision you reached changed when the only thing that changed between the two versions was a detail that should not bear on it, where a person works, how they speak, their age or caring responsibilities, how they came into the company or an accommodation they use. It does not measure prejudice, attitudes, beliefs, values, character or intentions, and it attributes none of them to you. It does not predict whether you will discriminate. It is not an implicit-association measure and is never to be described as or compared to one. It does not certify you or any organisation as fair or unbiased. It does not measure hiring decisions by interviewers, which is the Interviewer Judgment and Hiring Bias assessment; cross-cultural leadership, which is the Cross-Cultural and Inclusive Leadership assessment; or accessible design, which is the Digital Accessibility and Inclusive Design Judgment assessment. It reports what your answers did across twelve matched pairs in one sitting, and beside them, on its own scale, twenty unmatched items on inclusive practice. SCORING DESIGN, C18, matched-trial within-person contrast: each pair tells the same workplace decision twice, with the same facts, the same four options in the same order, the same reading load and the same warranted answer, differing only in one construct-irrelevant detail about a person in it. The two halves sit at least seventeen items apart. For each of four rows, grouping the pairs by what the detail was, the matched quantity is cost = raw accuracy on the unmarked halves minus raw accuracy on the marked halves, over the pairs whose two halves were both answered. The difference is taken on the RAW scale: chance-correcting each half against its own declared marginal and subtracting would make the figure move with the respondent's level rather than with the thing measured, because the two halves have different nulls by construction. That difference IS the ordinary cost, chance(unmarked) minus chance(marked), computed from the declared per-option marginals and printed beside the reader's own difference as the reference mark. A pair with either half unanswered leaves numerator, denominator and reference together. An empty sitting scores exactly zero on every row and is not reportable. Each row carries a standard error computed from the declared marginals as the standard error of a mean of its answered pair differences, a 68 and a 95 per cent band, and a verdict: the warranted decision was reached more often with the detail absent when the whole 95 per cent band sits above the ordinary mark, more often with the detail present when it sits below, and otherwise NO DIFFERENCE DETECTABLE IN THIS SITTING, printed at full size because on this instrument it is the commonest and the most important result. Omega per row is McDonald's omega in Spearman-Brown form from the answered pairs and an assumed inter-item correlation of .25, computed unrounded; with three pairs it sits under .70, so no row carries a number, and the refusal is printed in the reliability table. THERE IS NO COMPOSITE: no bias score, no index, no label, no type. Where a total would be, the report prints the refusal and two opposite authored patterns whose unweighted means are identical. Rows are ordered by the size of the departure from the ordinary mark in standard errors, never by level. The twenty unmatched items are scored as a chance-corrected quality against their declared marginals on their own scale, beside the rows and never inside them, carrying a figure only with at least eight answered and omega at or above .70. REPORT DESIGN, D21, the Conditions Board, adapted: each row is a horizontal bar between two named poles with the ordinary mark and the neutral zone drawn, and every row carries the sentence that this describes what your answers did across these matched pairs in one sitting, not who you are. Beneath the board every pair is reproduced with its two versions side by side, what differed and what you chose in each. SOURCES: Uhlmann and Cohen, Constructed criteria: redefining merit to justify discrimination (Psychological Science, 2005), for the criterion-first practice keyed on the pairs and the select-all; Bohnet, van Geen and Bazerman, When performance trumps gender bias: joint versus separate evaluation (Management Science, 2016), for side-by-side comparison; Correll, Benard and Paik, Getting a job: is there a motherhood penalty? (American Journal of Sociology, 2007), and Heilman and Okimoto, Motherhood: a potential source of bias in employment decisions (Journal of Applied Psychology, 2008), for the ask-rather-than-decide practice on the caring pairs; Posthuma and Campion, Age stereotypes in the workplace (Journal of Management, 2009), for the age pairs; Lev-Ari and Keysar, Why don't we believe non-native speakers? The influence of accent on credibility (Journal of Experimental Social Psychology, 2010), and Gluszek and Dovidio, The way they speak (Personality and Social Psychology Review, 2010), for the accent and language pairs; Rivera, Hiring as cultural matching (American Sociological Review, 2012), for the college and route pairs; Colella and Bruyere, Disability and employment (APA Handbook of Industrial and Organizational Psychology, 2011), for the accommodation pair; Hinds and Kiesler (eds.), Distributed Work (MIT Press, 2002), and Cramton, The mutual knowledge problem and its consequences for dispersed collaboration (Organization Science, 2001), for the site, remote and time-zone pairs and the rotation items; Karpowitz and Mendelberg, The Silent Sex (Princeton University Press, 2014), for speaking time and who is heard; Edmondson, Psychological safety and learning behavior in work teams (Administrative Science Quarterly, 1999), for the talked-over items; Czopp, Monteith and Mark, Standing up for a change: reducing bias through interpersonal confrontation (Journal of Personality and Social Psychology, 2006), for the flagged-remark item; Forscher, Lai, Axt, Ebersole, Herman, Devine and Nosek, A meta-analysis of procedures to change implicit measures (Journal of Personality and Social Psychology, 2019), and Paluck, Porat, Clark and Green, Prejudice reduction: progress and challenges (Annual Review of Psychology, 2021), for the true-or-false claim about one-off training and for the decision that this instrument reports decisions, not attitudes; Kahneman, Sibony and Sunstein, Noise (2021), for within-person inconsistency as the measured quantity; Hattie and Timperley, The power of feedback (Review of Educational Research, 2007), for the named-sentences practice on the proposal pair; Gollwitzer and Sheeran, Implementation intentions and goal achievement: a meta-analysis (Advances in Experimental Social Psychology, 2006), for the if-then form of the one change; Haladyna, Downing and Rodriguez, A review of multiple-choice item-writing guidelines (Applied Measurement in Education, 2002), for cue control. All forty-four items are original works written for this instrument; no item from any published questionnaire, implicit-association procedure, training programme or commercial inclusion instrument is reproduced or adapted, no branded instrument or vendor product is named or used, no competitor is named anywhere, and the instrument is not affiliated with or endorsed by any author or publisher named above.