Issue 53

Essential frameworks in agility

Gelu Vac
Software Engineering Manager @ Crossover


Despite everybody's good intentions, seemingly ignoring the effort invested and the relevance of the ideas thrown into the game, extremly lots of projects end-up creating products that nobody wants, nobody asked for and/or cannot possibly be sold. So, which might be the causes for such situations, considering that we live in an increasingly measured society, over-exposed to management processes and formal models, which practically guarantees minimum exposure to risk, from the success rate point of view?

Psychology economics

The dictionary defines "economy" as a "social science that is mostly interested with the description and analysis of production, distribution and consumption of goods and services ".

Another interesting definition of economy is "the study of the way people make decisions in situations of limited/scarce resources". This definition is extremely suitable for most branches of the classical economy.

This being said, macroeconomics is the study of decisions that people make under the pressure of limited/scarce resources at a national or global level. It deals with the effects of the decisions made by nation leaders, in terms of taxation value, interest amount, and external and trade politics.

On the other hand, microeconomics is the study of the decisions people make under the pressure of limited/scarce resources at a personal level. It deals with decisions made by individuals and organisations about exposure to risk, the gear they should invest in or establishing prices for the products or services they have to offer.

Like game theory, psychological economics clearly expands the range of phenomena economists can successfully study, and does so in what clearly is the spirit of economics. And as I discuss below, like game theory, psychological economics is destined to be absorbed within economics, not exist as an alternative approach.

How can economists embrace all that is good in economics while dedicating ourselves to a more realistic conception of human nature, as it pertains to economic situations? The methods of economic analysis—methodological individualism, mathematical formalization of assumptions, logical analysis of the consequences of those assumptions, and sophisticated experimental and field-empirical testing have many virtues. But these methods create a necessary evil: we must use highly simplified and stylized models of human cognition.

Now let's focus on a couple of well-known and often used models, in the contemporary economic and industrial ecosystem.

Utility functions

It is an important concept that measures preferences over a set of goods and services. Because satisfaction or welfare is a highly abstract concept, economists measure utility in terms of revealed preferences, by observing consumer choices and creating an order for consumption baskets from least desired to the most preferred.

Today's utility functions, expressing utility as a function of the amounts of the various goods consumed, are treated as either cardinal or ordinal, depending on whether they are or are not interpreted as providing more information than simply the rank ordering of preferences over bundles of goods, such as information regarding preference strength.

Behavioral economics

This field studies the effects of psychological, social, cognitive, and emotional factors on the economic decisions of individuals and institutions, and the consequences it bears on market prices, returns, and resource allocation. The field also tackles, not always that narrowly, but rather more generally, the impact of different kinds of behavior, in different environments of varying experimental values.

The framework for modifying unrealistic assumptions

There are many assumptions that economists often make about human nature that behavioral and psychological research suggests are often importantly wrong. These include the assumptions that people:

The goal of psychological economics is to investigate behaviorally-grounded developments from these assumptions that seem economically relevant. For a more concrete frame of reference, consider the following formulation of the classical economic model of individual choice, where uncertainty is integrated as probabilistic states of the world, with a utility function that may depend on these states of the world, and the assumption that the person maximizes expected value:

where X is choice set, S is state space, π(s) are the person's subjective beliefs updated by Bayes' Rule, and U are stable, well-defined preferences. From this characterization of the "classical" model, I like to categorize psych phenomena for economists into three categories:

  1. New assumptions about preferences - what does U(x|s) really look like?

  2. Heuristics and biases in judgment - how do people really form beliefs p(s)?

  3. Lack of "stable utility maximization" - do people really follow the Max function?

These goals involve trying to be as clear and orderly as possible in identifying exactly what developments are necessary, meaning that organizing these developments could identify as precisely as possible where and how classical economic assumptions go awry. As always, the tension between clarity/conceptual tightness versus truthfulness to the behavioral and psychological reality is a core problem in economic modeling.

The first category of developments is to identify ways to make U(x|s) more realistic, while maintaining the assumptions that beliefs π(s) are formed rationally and that people fully rationally maximize the given equation. Examples include the assumption that people have reference-based utility - they care a lot about changes (in wealth, consumption, ownership, health, etc.), not solely about absolute levels. It also includes non-expected utility - that preferences are not linear in π, as in the above formula, but rather maximize a more general form U(x,π). People care differently about uncertainties reflecting subjective "uncertainty" and those uncertainties reflecting objective "risk".

The second category of developments consists of ways by which people form potentially distorted beliefs p(s) about the world rather than form beliefs π(s) through proper Bayesian reasoning. Research on judgment under uncertainty identifies the role of heuristics and biases in forming probablilistic beliefs.

The third category of assumption modification is to consider psychological findings that suggest that there may not be stable, well-defined, time-invariant, and "hedonically correct" preferences U(x|s) such that behavior is best described by assuming that people maximize . Examples here include exploring the ways that people mispredict or misremember their own utility—there are identifiable patterns in how people misperceive their own future taste (e.g., they under-estimate how much those tastes will change), and even in how they evaluate their experienced well-being from past episodes (e.g., they tend to under-emphasize the duration of the episode).

Diminishing-marginal-utility-of-wealth framework

There really doesn't exist a non-insane utility function within the expected-utility, such that you will turn down $100/$110 bets over a broad range of initial wealth levels. Any such utility function also predicts outrageously wild risk aversion over larger stakes.

Losing $20 in a bet, losing a watch we just bought, or losing $10,000 in the stock market in a week all feel bad, but tend to feel worse because we too rarely think in the broader, long-term perspective, where these losses will almost surely be wiped out by other gains in the long run.

Caring about others

Economic actors are mostly self-interested. But as many economists have recognized over the years, self-interest, as narrowly defined in virtually all economic models, isn't all of human motivation. Moreover, the departures from the standard self-interest assumption are potentially important for economics, for issues like understanding the short-run reaction to market price changes, political economy, and special labour-market institutions.

A simple hypothesis for how people care about others' well-being is natural for economists, and has the longest history in economics: Altruism - positive concern for others as well as yourself. Altruism can be either general or targeted; you may care about all others' well-being, or maybe selected others' (friends, family) well-being.

But such simple altruism is not adequate for understanding many behaviors. Two other aspects of social preferences show up prominently in psychological and recent experimental-economic evidence. First, people care about the fairness and equity of the distribution of resources, beyond ways that increase the total direct well-being. Second, people care about intentions and motives, and want to reciprocate the good or bad behavior of others.

Expected-utility framework

The classical expected-utility framework predicts essentially risk neutrality over non-huge stakes. Our attitudes towards risk are instead primarily influenced by the attitudes towards change in wealth levels.

The way that people isolate separate instances of monetary gains and losses relates to a major problem in economics. Perhaps the most often used assumption in economics is that "risk aversion" derives from diminishing marginal utility of wealth within the expected-utility model: U"(w) < 0.

If (say) you dislike a 50/50 Lose $100/Gain $110 gamble, it is not because of the change in marginal value of consumption due to $100 decrease or $110 increase in your lifetime wealth. This is simply way too big a change in marginal utility for way too small a change in wealth.

Your current life time wealth is a complicated stochastic creature with some huge mean and variance. And your reaction to a loss of $100 isn't the difference (in your anticipated lifetime expected utility) between your existing complicated stochastic distribution of lifetime wealth and your new distribution of lifetime wealth corresponding to the shift $100 to the left of this big complicated distribution. Rather, what is salient to you is your sensation of losing $100.

Reference-dependence can be thought of simply in terms of a new assumption about preferences. Letting c be a vector of the levels of wealth or consumption of goods or activities, and r be a vector of reference levels in the same dimensions, incorporating reference dependence into utility theory involves merely a switch from the function U(c) to U(c,r).

Care about change

A core feature of humans is that we are highly attuned to changes in our circumstances, not merely the absolute levels. We can feel colder, even in the same attire, if it is 10 degrees in the summer rather than if it is 7 degres in the winter.

This fact about human nature carries over preferences. For instance, our sense of well-being from our total consumption is not solely a function of its level. It is a function of how that level compares to what we are used to. The related phenomenon of hedonic adaptation is a primary fact about human nature: Even for major life events, once a new steady state is reached, we tend to return to the previous hedonic level over time. So, the event of becoming wealthy, not just being wealthy, can often be a major source of satisfaction, and once we get used to a new standard of living we may, day to day, be roughly as happy as when we were poor.

While the identification and measurement of how we feel about changes is an active area of research, one core aspect of our reference-based preferences is known to be crucial: loss aversion. The sensation of loss relative to status quo and other reference points looms very large relative to gains.

Utility-maximization framework

People rejecting unfair offers in the ultimatum game are consciously choosing, among two possible outcomes, the one they prefer.

As an economics concept, it focuses on maximizing the total value derived from the available money. When making a purchase decision, a consumer attempts to get the greatest value possible from the expenditure of the least amount of money.

Care about self-interest

Departures from self-interest is largely compatible with the utility-maximization framework.

To many of us, seeing an experimental subject sacrificing (say) $8 to punish an unfair ($92,$8) offer in the ultimatum game looks straightforwardly like a preference for the allocation ($0,$0) over ($92,$8) when motivated by retaliation. There is simply nothing perplexing about somebody sacrificing $8 to punish a jerk who wants to split $100 $92/$8 rather than $50/$50 (or at least $60/$40).

Exponential discounting framework

To model intertemporal preferences formally, let Uτ be "intertemporal preferences" and let ut be instantaneous utilities. Economists assume exponential discounting:

, where r > 0 is a parameter. The first alternative to exponential discounting proposed by psychologists and others trying to capture presentbiased preferences was "hyperbolic" discounting:

, where k > 0 is a parameter. In part, because the continuous-time hyperbolic discounting function is difficult to deal with, and, in part, because the specific functional form of hyperbolic discounting is neither literally correct nor very important, researchers have recently modelled present-time biased preferences with the following discrete-time discounting function:

for all t, , where the parameters β and δ are less than 1, with δ very close to 1. The discrete-time exponential model corresponds to β = 1.

Care about now

People like to experience pleasant things soon and to delay unpleasant things until later. To capture this preference for gratification earlier rather than later, economists traditionally model such tastes by assuming that people discount streams of utility over time exponentially.

But the exponential form of discounting has a special property that has been shown repeatedly to be false. It is the unique functional form that generates time-consistent preferences, whereby the preference between any two intertemporal trade-offs in momentary well-being - between, say, getting lesser satisfaction earlier versus a greater amount of satisfaction later—is the same no matter when asked. The behavioral evidence, by contrast, overwhelmingly and incontrovertibly shows that people exhibit present-biased preferences.

Consider, for instance, the following two choices of work patterns:

Suppose that opportunity costs of time, the disutility of work, the productivity of work, etc., are all identical on April 1 versus April 2.

If asked to make the choice above, on January 1, you will surely prefer the first to the second choice, since it involves less work in total. However, if asked on April 1, you might choose the second because, if asked on April 1, the choice is now between 7.0 hours of work today and 7.7 hours of work tomorrow.

How are these preferences time-inconsistent? They would predict precisely the behavior discussed in the work example above, for instance, if we set β = 0.8 and δ ≈ 1, if we assumed the disutility of work is linear in hours worked. 7.7 hours of work 91 days from now generates perceived disutility of 0.8 x 7.7 ≈ 6.2, which is less than the perceived disutility of 7.0 today.

Suppose [u(relax)-u(7.7 hours work)] ≥ 1.1 [u(relax)-u(7 hours work)]. That is, merely assume that there is a non-decreasing disutility of work. If you try to explain preferring 7.7 hours of work tomorrow to 7 hours today with exponential discounting, then you must assume a yearly discount factor of δ < 0.00000000000000002. This conclusion comes from the arithmetic truth that 0,9365 ≈ 2 x 10-17. Discounting by 10% from one day to the next means that over a year you will discount at the rate of 0,9365 - if you assume, as you must if you believe in exponential discounting, that the discounting will be at the same rate level each day . Since this is a ridiculous - and behaviorally counterfactual - discount factor, we know that the observed discounting is not consistent with exponential discounting.

By other similarly-easy arithmetic exercises - such as observing that 0,99365 < 0,026, 0,9931 < 0,74, and 0,999365 < 0,7 - we notice an even far less extreme taste for immediate gratification.


For any natural action, there is a purpose and a natural context which led to its manifestation. We will name this context as framework, and we will assign it the responsibility of reaching that purpose.

A product is not being delivered by a man or a team, but by an entire gear. In order to have a happy Customer, agility has to manifest in absolutely every single phase of interaction with the Customer: from negotiation, offer and contracting until delivery, waranty and post waranty services. Moreover, the project budget is not a variable of that project, but one of the constraints. The same goes for the deadline.

Agility is a framework meant to reduce the level of formalization and NOT another formal model to manage a production flow.

A manager has to behave as the aggregator of the team, as a fundamental reference in terms of behaviour. The manager is the role-model. He is not the "higher in rank" member of the team, but the person responsible for both the success and failure of the team.

Performance evaluation should never be applied on small intervals (daily or weekly), running in deep collision with the main focus of the members of the team, but rather make use of long term vision and focus on continuous increase of motivational incentives for team members.

Candidates' evaluation needs to go beyond technical assesment and touch a cultural component, attuned to the already created micro-climate inside the existing team. Relationships with team members have to be equidistant, beyond personal preferences and interests.

Involvement in any project should be completely altruistic: we do not deliver projects for ourselves, but for the Customer who placed the initial order and budgeted the effort. We can and have the professional duty to invest innovation and creativity in projects, but within the critical constraints of Customer-utility and preference.

Decisions taken inside a team have to be based on reciprocity - any decision produces effects which have to be assumed by the entire team.




  • Accenture
  • BT Code Crafters
  • Accesa
  • Bosch
  • Betfair
  • MHP
  • Connatix
  • BoatyardX
  • Telenav
  • .msg systems
  • Grab
  • Yardi
  • Colors in projects