Can psychological safety be measured? How trita estimates it
How safe is it on your team to ask, to admit a mistake, to disagree? Here is how trita estimates it - and what to do with the index as a leader.
In this article
Psychological safety can be estimated, but only from team members' self-reports. Amy Edmondson's 1999 paper introduced the concept and the seven-statement, team-level scale that goes with it. Trita's eight-statement short assessment starts from this concept. The result measures perception, not the objective quality of a team.
What does the concept mean?
Psychological safety expresses how safe team members feel about asking questions, admitting mistakes, asking for help, and voicing a differing professional opinion. It does not mean comfort. Nor does it mean the absence of conflict.
In a psychologically safe team, professional disagreements surface more openly, because members have to spend less energy on self-protection.
In practice you see it in who dares to say in a retrospective that a decision was wrong.
According to the 2017 meta-analysis by Frazier and colleagues, psychological safety is associated with learning behaviour, information sharing, and speaking up. It is also related to performance, though more weakly. It is an important team factor. Citing it as the universally strongest predictor would be an overstatement.
What did Google's research show?
Google's internal analysis, called Project Aristotle, examined 180 of its own teams. Among the five dynamics associated with effective working, psychological safety was highlighted first. The results were made public in 2015 on the re:Work site.
This is a valuable organisational case study. It is an internal, observational analysis, so it does not automatically generalise to every company. The composition of the teams and the organisational context were specific too.
Nor does it follow from the research that individual knowledge is irrelevant. The lesson is rather that without the quality of interactions, even excellent individual capabilities are not necessarily put to use. These two statements are not the same.
How does trita estimate it?
Edmondson's original seven-statement scale examines how members perceive the handling of mistakes, difficult topics, asking for help, and interpersonal risk. The measurement is interpretable at team level, because the concept concerns shared perception.
Trita's own short assessment consists of eight statements. It is not identical to the original scale. To call it an independently validated instrument, we would have to document the origin of the items, the reliability, the factor structure, the aggregability to team level, and the sensitivity to repeated measurement. We are working on that documentation, and we will publish the results on this blog.
Our index gives an estimate of team members' perceived psychological safety. It passes no objective judgement on the team or on the leader. This is precisely the hardest thing to get a direct picture of as a leader, since someone who does not feel safe speaking up is unlikely to say so out loud either.
This is what the psychological safety slice of the report looks like:
Psychological safety pulse
Team index and areas – without individual answers
7 responses · The index only appears from 3 responses.
Alongside the team index, the average per area is visible too, with the weakest area highlighted. Individual answers appear nowhere.
What does the threshold of three mean?
We handle the responses to the short assessment without direct identifiers. Below three responses we display no team index. This reduces the risk of directly revealing individual answers.
Three responses is not a scientifically established anonymity threshold. In a small team, wording, timing, or circumstances can still make it possible to infer who said what. That is why we speak of confidential handling without identifiers and of a minimum reporting threshold. We do not promise full anonymity.
For credible answers, participants have to understand how the data is handled, and they must not fear personal consequences. This is a necessary condition. It does not on its own guarantee the validity of the measurement.
From the index to the leadership conversation
The index is not an instruction to a leader. In a consulting conversation it can be set against the team's own experience, other aggregated data, and the actual work processes. That is how it offers a perspective for the next conversation.
Four typical experiments tend to follow:
- Reviewing mistakes for learning. In the retrospective you first look at what you learned from a mistake, and only then at how to fix it.
- Voicing leaderly uncertainty. An „I am not sure about this, what do you see?” opens a conversation better than a ready answer.
- Asking for the minority view. Before a decision, you ask by name the person who has not yet spoken.
- Making it easier to ask for help. Give it a stated channel and time, so it is not a test of courage.
These are not recipes. They are experiments adapted to the team's situation, worth reviewing after a few weeks.
When can you speak of change?
A second measurement shows a difference in the index. To read that as a meaningful change, you have to take into account measurement error, who responded, the team's composition, and what happened in the meantime. One departure and one arrival can shift the average on their own.
That is why in our comparison report the measurement uncertainty and the change in composition are part of the interpretation alongside the point difference. If the difference cannot be separated from the uncertainty, we do not claim that the team's psychological safety improved or deteriorated.
What not to do with the index
The index attracts a few characteristic mistakes. One is measuring teams against each other. A lower value compared with other teams in the organisation says nothing about the leader, because the task, the workload, and the team's history all differ.
The other mistake is turning it into a target figure. If the index becomes a leadership objective, the next round of responses will be distorted. Participants know what is expected of them, and the measurement loses its value. So do not tie the index to bonuses or reviews.
Limitations and sources
Our eight-item short assessment is our own product solution. The literature cited supports the conceptual basis. It does not substitute for the psychometric validation of our own scale, and we have not completed that validation.
If you are planning the introduction, read our article on introducing an assessment first. The voluntariness of a measurement directly affects how credible the results are.
- Edmondson (1999): Psychological Safety and Learning Behavior in Work Teams
- Frazier et al. (2017): a meta-analytic review of psychological safety
- Google re:Work: the method and results of Project Aristotle
- Newman et al. (2017): a research review of psychological safety
A note on the references: Google published the Project Aristotle results in stages between 2015 and 2016. The year 2015 refers to the first public appearance of the re:Work guide, which is worth checking if you cite an exact date.
Trita's eight-item short assessment is our own product solution. The literature above supports the conceptual basis; we document the psychometric properties of our own scale publicly, as data accumulates.
See what your results look like.
In ~10 minutes, get a useful picture of what you bring to a team – for free.
Értesítő
Új cikkek és válogatott összefoglalók
Szólunk, ha új cikk jelenik meg, és időnként hírlevelet is küldünk válogatott olvasnivalókkal. Bármikor leiratkozhatsz.