Fox on Software, by @sleepyfox on 21 Jan 2014
"Ah, now we see the violence inherent in the system!"
I hope that this provides the reader with some food for thought, and encourages your curiosity enough to ask some of the more challenging questions around metrics, measurement and quantification, both at work and in your broader life.
NVC, or Non-Violent Communication, is a way of communicating invented by Marshall Rosenburg as a way of facilitating and mediating between parties in conflict resolution scenarios between civil rights activists and the administration in 1960s America.
Non-violent Communication asserts that most conflicts between individuals or groups arise from miscommunication about their human needs, due to coercive or manipulative language, these "violent" modes of communication, when used during a conflict, divert the attention of the participants away from clarifying their needs, their feelings, their perceptions, and their requests, thus perpetuating the conflict.
NVC is based on the idea that all human beings have the capacity for compassion and only resort to violence or behaviour that harms others when they don't recognize more effective strategies for meeting needs.
One aspect of NVC that was new to me when I was first introduced to it by Bob Marshall was the idea that a style of communication could be “violent” in a way that was not overtly aggressive or laden with profanity.
Instead he showed the room that there are many ways that language can be 'violent', for example:
- Judgement - blame, comparisons, criticisms, insults, labels "You are so wrong!"
- False feelings - belief that you are being made to feel something "You are making me angry!"
- Should - belief that the other person must do something "You should tidy up now!"
- Demands - use of coercion - "and you better, or else!"
In a recent Twitter conversation Bob asked me to explain what I meant when I suggested that quantification was, or even could be, violent.
The conversation had strayed into the area of Measurement, Quantification and Tom Gilb. For those who are newcomers to Gilb's Evolutionary Management, which according to Gilb "makes project failure structurally impossible!" (his emphasis, not mine), a bold claim.
One of the central tenets of 'Evo' is that everything can be measured, and "Anything you need to quantify can be measured in some way that is superior to not measuring it at all." - see here.
People in 'Evo' seminars come up with a way of measuring the love of their spouse, partner or family member, quantifying it by things like the number of presents given, the amount of time spent together, how many times they say "I love you" etc. The idea is that you take these measurements, apply a weighting factor and use them to derive a metric that gives you an overall "Love index" that you can then use as a baseline for changes in their (or your) behaviour.
There are many many problems with the "measure all the things" approach, and I don't mean the typical excuses such as:
- "This is too hard to measure"
- "This is too costly to measure"
- "Measuring will take too long"
- "We don't know enough about this to put numbers on it"
Gilb himself says that these arguments are irrelevant, because most of the gain of a metric is to be had from the understanding of a thing that is gained by deciding how it can be measured, which he calls 'Quantification' (literally: 'the assigning of a number to a thing') - see here. This is analogous to Eisenhower's "plans are useless but planning is indispensable" quote. So let's go with the grain and just focus on quantification, rather than measurement, leaving aside the issue of:
- Considering Kaner's Ten Measurement Factors, what are the likely variances and confidence intervals in my metric?
Now what we're talking about are some of the more fundamental issues such as:
- How might we quantify this thing?
- What is it we hope to gain (by quantifying this thing)?
- Which parties stand to gain most?
- What is it we stand to lose?
- Which parties stand to lose most?
- What possible side-effects are there?
- How might parties “game” or manipulate this metric?
- What biases are present?
- How can we offset or mitigate those biases?
I'm not even going to talk about these very real issues, which are in my experience almost never considered, but rather focus on the even more important issues of:
- For the decision to quantify a thing to make sense, the underlying entity must be amenable to measurement at a level and reliability that makes deducing information from this measurement feasible.
- The choice of a quantification scheme for this thing is inherently the imposition of a particular world view or schema on this thing, complete with that world view's value judgements and presuppositions.
The first is a particular problem with algorithmic complexity, which manifests itself in that old software bugbear: estimates, and I talk about this in my presentations on Software Metrics e.g. here explaining the very real problems that this causes in software development.
But it is the second problem that I tweeted about, that in choosing a quantification scheme for a thing, I am in fact imposing my world view upon that thing, and whilst that may seem an obvious and trivial statement to many, it can have profound consequences.
In my experience, the choice of how a thing is quantified, measured, and how a metric is applied is almost always done by a single person (or a small clique), and those to whom the metric is applied are neither asked whether they wish to be measured nor consulted on what might the best or most appropriate means be to measure them. Their views are unasked for and unheard. For example consider a new performance review process at work being rolled-out by HR. Are you asked how you feel about your performance being measured, and in this way? Are you consulted on how effective measurement of your work might be done? How does this make you feel? Does it make you feel valued? Trusted? Listened to? How might you react to this measurement scheme? For a real world example of how this can go horribly wrong see Microsoft's stack-ranking system, e.g. here
How would you feel if your partner said that they were going to be assessing your relationship by using the "Love metric" that we talked about earlier? Would it make you feel more loved? What kind of behaviour changes might that encourage?
I suspect that this very effect is one of the root causes of the appalling "70% of workers remain disengaged by their work" statistic, see Gallup poll from last year here. In case you thought that being unhappy at work isn't really such a big deal, Gallup calculates that this costs the US economy alone $350 billion. Per year! In UK terms taking the figure per capita that is equivalent to 10% of the national deficit!
This is the imposition of one's will, world view and value system on others, and to my eyes this clearly falls under the NVC definition of violence, incorporating elements of Judgement, Should and Demands.
To be clear: I am not saying that quantification - or measurement or a metrics system - is inherently violent, but only that the way that it is most commonly implemented is. This is analogous to saying that the English language is not inherently violent, but the way that it is most commonly used is.
What do you think? What quantification or measurement schemes that impact you in your day-to-day life would you like to see improved?