This week we discuss the topic, ‘Why do big projects fail’. With a slant towards IT projects due to our backgrounds in the field, we discuss and explore some examples of failures to date.
There are different types of project failure, where the project is: complete but costs too much; complete but too late; complete but with low quality; incomplete and doesn’t do what was originally envisaged; combinations of the four; and not done at all – scrapped by the stakeholders with nothing to show for it.
Software and IT projects, more than other types of projects such as construction or manufacturing, often suffer from the last case – scrapped. We discuss agile methodologies and whether they are the solution – or at least whether they reduce the probability of failure.
Sometimes there are too many chiefs and not enough engineers. And even if the business analysts identify the full scope of what needs to be done, sometimes the stakeholders don’t see the same vision or can’t afford the required budget.
Everyone makes mistakes so why can’t you? – Big Bird, Sesame Street
Life is all about making mistakes. As we age, interrogating our own personal histories for mistakes yields more and more useful lessons, for those who wish to go through the exercise (it’s painful to do so). Whether it is in a performance review at work, on stage, in sport or just a quiet reflective moment at home, there is nothing quite as fruitful as constructively inspecting your mistakes and learning from them.
Mistakes are information wrapped in pain. They seem to be very good at compressing a lot of information into a small package. If you think of adaptation and evolution as an information processing system, then mistakes (errors, evolutionary dead-ends) are the primary way to glean information about the environment for the system.
In tonight’s discussion, we start with personal anecdotes that show the same quality – mistakes we have made that have become touchpoints for our own lives. Mistakes which have become something much bigger and much more positive.
What would the world be like if no one ever made a mistake? What would a universe without error look like? Would evolutionary processes grind to a halt? Could we ever have got out of the primordial soup without our propensity to make errors?
But then again what is a mistake? If something was a good idea “at the time”, it only become a mistake later in time. This binary judgement (mistake/not a mistake) cannot be applied to the event, but only to the complete system in time and space (action + time + consequence). And even then, whether or not something was ‘wrong’ depends a lot on your frame of reference as an observer.
Our culture is schizophrenic about mistakes. We pay lip service to the ancient ideas of empiricism and learning by “trial and error”, but our world is theory driven and in love with optimization, which implies the removal of redundancy and error. Our default position is to design our lives in such a way as to minimize our opportunities for making mistakes.
But you have to do it wrong to do it right. To learn what not to do. A life/world without error is dead, boring and sterile.
As Gentlemen Scientists we have made more than their share of mistakes. But mistakes plus attention equal learning. Perhaps a good life is not making less mistakes but always making new mistakes. Plus being aware of your fallibility – having the courage to say “I was/could be wrong”.
3 – Impostor syndrome – a psychological phenomenon in which people are unable to internalize their accomplishments. Despite external evidence of their competence, those with the syndrome remain convinced that they are frauds.
4 – I love this anecdote from Prosunjit Biswas about cooking a Bengali dish called ‘mishti doi’ – he had to get it wrong in a high profile situation before he really learnt what not to do.
5Dennett says here that “sometimes you don’t just want to risk making mistakes; you actually want to make them — if only to give you something clear and detailed to fix.”
8Imagine a spider catching flies. If the spider always gets his prey and never fails to catch a fly, would that mean that spiders flourish or would that mean that eventually spiders die out. Do we need mistakes to grow and flourish. Are errors the food of life.
[Today’s discussion is about Constructor Theory as proposed by David Deutsch. The original 2012 paper can be found here .]
A recent article in Scientific American introduced us to the ideas of David Deutsch and his Constructor Theory. Constructor Theory seeks to formulate “meta-laws” which sit “above” the laws of physics and determine them. They would do this by describing what may and may not (i.e. is forbidden to) happen, rather than trying to explain what will happen. Deutsch proposes that such a “meta” framework may be the way to unify the quantum and classical models of physics.
Deutsch is known as a pioneer of Quantum Computation, but this theory of his is speculative and not well known. It has its detractors, such as a this particularly abusive young physicist from the Czech Republic. But Deutsch’s ideas got us thinking here at Gentlemen Scientists HQ and set us off on a rambling, semi-informed conversation.
The thing is, we love the “meta”. Meta laws are fascinating because they hold out the promise of deeper understanding and unification. Understanding and control at a meta-level is stable over time than first-level control and survives the unexpected. Meta cognition, for example – the ability to reflect on one’s own patterns of thinking and behaviour – leads to better strategies for living under conditions of uncertainty.
However, once you start up the “meta-” ladder, how do you stop? Are there meta-meta-laws, or meta-meta-meta-laws? And when you do stop, you’re still left with the question – who designed the metaN law that you are left with? You’re still left looking at the “Face of God”.
So this approach is no magic bullet. In a way, the Constructor Theory is a deliberate lowering of our ambitions. Our tendency is to want to predict, and we judge our scientific models by their predictive power. But if we shift our focus from what will happen to simply trying to write down what will never happen, and then following such an exercise to its logical conclusion, we may be surprised to find just how much “drops out”(this seems to be what happened with Deutsch – he ended up with results that a remniscent of open problems in quantum mechanics).
If it makes us feel better, we can say that we are “deferring” the question of working out the “laws” while we explore the “meta-laws” first. Will it turn out that we never actually need to go to the “laws” in the end, because the “meta-” understanding will give us what we need? Counter-intuitively we might get a deeper understanding by trying to understand “less”.
Slight tangent: it got us thinking about our cognitive biases. When we look at network graphs, we concentrate on the nodes, not the edges. We put people as records into our databases, but not the relationships between people4. We see things, and we try to describe the things, but we don’t think as much about the things that make things or the things that connect things.
Crude analogy with a network graph. Usually we like to think about the nodes N of the graph, but for a connected graph we could work with the edges A only and ignore the nodes altogether without losing any clarity or completeness of our model. We could even derive second order measures B based on the edges, such as e.g. relative sizes of adjacent edges. The resulting set of “meta”-edges would no longer fully specify a single graph but a whole family (“universe”) of possible graphs.
Before we get too far into our own navels, however, we do end up with a couple of concrete lines of enquiry. Deutsch’s work may be crackpot, speculative or revolutionary – or anywhere in between – but we feel encouraged to follow the same process when thinking about people and memes. Centuries of efforts to codify laws about people and communication have met with limited success in our opinion (sorry, social scientists).
So is it time to take a meta view and develop some meta-laws about what people don’t do instead? For example, we could imagine that a simple biological imperative make it (almost) impossible for us to truly believe in the imminence of our own death. What happens if we represent some of these basic assumption and construct ‘laws” from them? Could we construct a better social science that way? We’re excited about that possibility and are thinking of some simple computer simulations to explore these ideas further.
4I had a recent conversation with a database expert and data archivist where we discussed the fact that we rarely design databases to represent the relationships between people, instead representing the people themselves. Kind of like representing the nodes of a graph but ignoring the edges. Although I am sure that sites such as LinkedIn have data structures that are doing that in some way. I wonder what would happen if we completely (and counter-intuitively) “ignored” the people themselves and only represented the relationships. People would still exist but only as an “inferred” property of the system.