How People Contribute to Growth-curves
How People Contribute to Growth-curves
Prof. Peter P. Robertson, M.D.
Affiliations: Monterey Institute of International Studies, Monterey, USA
Prof. dr. Wouter Schoonman
Affiliations: Psy Tech industrial psychology, Den Haag, The Netherlands and Saxion and
Rotterdam Universities of Applied Sciences, The Netherlands
This is a working paper; version 2.2
Abstract
Research started originally at McKinsey and Company (Thomson, 2006) into success
factors of about 1700 Initial Public Offerings (IPO’s). In many of the 70 companies in this
research project that made it to a USD billion turnover, personality patterns in the founding teams
could be positioned on opposite positions of the Growth-curve.
This research made use of an ecological tool, the AEM-cube® (Robertson, 2005). This tool
relates three key characteristics of the contribution of personalities to Growth-curves: first to
what phase of a Growth-curve a personality contributes, second whether the personality is
attached more to either technological Growth-curves or commercial Growth-curves and third
whether a person is focused on a specific part of a Growth-curve or is focused on integration of
larger parts or even the whole of a Growth-curve.
This approach makes it possible to construct Growth-curves by aligned personalities in a
relay kind of sequence, matching their specific contributions to the successive phases of a
Growth-curve.
AEM-cube® perceptions generated by Silicon Valley based observers about how Steve Jobs
and Tim Cook aligned themselves to create a fruitful Growth-curve or the original context of the
founding team of Yahoo (Thomson, 2006) are used, amongst others, as examples.
Copyright © 2013 P. Robertson
Page 1 of 30
How People Contribute to Growth-curves
Given the results from the research mentioned above and many other similar case by case
examples of the relationship between the contribution of individuals and their appropriate relaylike alignment to Growth-curves, an obvious question is how stable these personality
characteristics are over time. If there is a long time stability of these characteristics, this approach
will open a route to a long term strategic human resources management, being able to create
optimal conditions for every phase of a Growth-curve encountered in the products, services or
client relationships in organizations. It could also create an approach for individual career choices
of individuals, matching their personalities with the types of functions, roles and assignments in
organizations.
This article describes the basic statistical background of the AEM-cube® and the
longitudinal research of all assessment and re-assessment data, within a time range between 1 and
12 years, that could be extracted from the 30.000+ assessments available today.
The result will show that there is a high level of stability and that assessment and reassessment data do not differ more than about 10 percentiles over the years for the two factors
that describe the direct contribution to the Growth-curve. The third factor, describing the
contribution to the integration of Growth-curves differs about 15-20 percentiles, which was to be
expected, because this factor reflects in a certain way a personal development as a consequence
of career development and was never hypothesized to be stable in the first place.
The conclusion is that the AEM-cube® can contribute to a long term strategic human
resources management both from the organization to optimize individual contributions to the
strategic phases of growth, as well as from individuals to organizations to optimize their
individual career paths.
Keywords: Organizational Ecology, Ethology, Cybernetics, Complexity-Theory, AEMcube, Growth-curve, S-curve
Copyright © 2013 P. Robertson
Page 2 of 30
How People Contribute to Growth-curves
Introduction
The research on success factors of about 1700 companies that went public, originated at
McKinsey and Company. The research was later continued by Thomson as an independent
consultant (Thomson, 2006). He found that in many of the 70 companies that made it to a
turnover of a billion USD, patterns of personalities within the founding teams could be positioned
on opposite positions of the Growth-curve.
David Thomson named this phenomenon ‘the outside-inside dialogue’, Charles O’Reilly
(Stanford) and Mike Tushman (Harvard) named a similar phenomenon ‘ambidextrous
management’ (O'Reilly, 2004) and Bob de Wit (Nyenrode) named the phenomenon ‘strategic
paradoxes’ (De Wit & Meyer, 2010).
An example used in Thomson’s book - Blueprint to a Billion - is the case of Yahoo.
Thomson recognized that in many of the successful founding teams, a typical pair of founder
personalities existed. He named them, rather intuitively, the Mr. Outside and the Mr. Inside and
he defined the dialogue between them as critical for starting a Growth-curve. He saw this InsideOutside duo as an asset for investors. Based upon the credibility the AEM-cube® tool received in
the Hewlett Packard/Compaq merger (Robertson, 2005), it was used to assess a series of
founding teams of those companies that ‘made it to a billion’ described in Thomson’s book and
supported his observations. Yahoo was one of them and the – typical - example is given below.
An assessment of the Yahoo founding team, like with most other teams in Thomson’s
research, was based upon the biographies and other data available about the founders and key
members of the founding teams. An independent group of observers, aware of the AEM-cube®
concepts and aware of the people involved, were asked to score the team members via a web
based questionnaire. Where possible, like in the Yahoo case, the results were checked against
observers who knew both the team members well enough as well as were proficient in the
understanding of the AEM-cube®.
The first answer this assessment delivers, is the position of the contribution of each
individual to a Growth-curve, also often called S-curve (Modis, 1998) .
Copyright © 2013 P. Robertson
Page 3 of 30
How People Contribute to Growth-curves
Figure 1 Key Founding Members of Yahoo
In the case of Yahoo, this profile shows that one of the technical founders, Yang, is
perceived like contributing close to the earliest stage of a Growth-curve. Filo (the other technical
founder) and Koogle (who joined as CEO to develop the company) are perceived at about the
first inclination point of the Growth-curve, which is the position where the real upscaling is to
take place. Koogle is known to be the builder of Yahoo.. Filo and Yang were together the
technological founders, for reasons that will become clear below, Mallet is positioned to
contribute as COO in the second half of the Growth-curve, where structuring activities like
control of any kind (like financial, legal, organizational design and quality management to name
a few) are the main contributions to make.
Edwards is also shown in this picture as she epitomizes with her profile the operational
contribution that is needed in the middle of any growth-process.
The key finding of this type of assessment, as illustrated by the example of the Yahoo team,
is that in the founding teams in Thomson’s research, team members were optimally aligned along
their contributions to a Growth-curve. In general, the dialogue between people who contribute to
the early stages as well as to the later stages of a growth-curve, like between Mallet and Koogle,
is called ‘the outside-in-inside-out’ dialogue and, more informal, because it is perceived so often
as effective, the ‘Golden Dialogue’ or ‘Mr/Mrs Outside - Mr/Mrs-Inside Dialogue’.
In order to expand this example to a broader insight in the dynamics of teams, it is
appropriate to first explain the model of the AEM-cube® as a whole.
Copyright © 2013 P. Robertson
Page 4 of 30
How People Contribute to Growth-curves
The AEM-cube® backgrounds
The AEM-cube® differs fundamentally from most existing psychological assessment tools,
by the falsifiability of its purely cybernetical foundations. This means that if one of the three
constituent concepts (paradigms or laws) it is based upon, would be proven to be false, the AEMmodel will be wrong also. Two of the three constituent concepts are the basic instinctive
biological systems, found in human beings and other social animals: ‘attachment’ and
‘exploration’ (Lorenz, 1981). The third constituent concept is The Law of Requisite Variety of
Ross W. Ashby (Ashby, 1956) who can be seen, with Norbert Wiener, as the founder of modern
cybernetics (Pickering, 2010). Cybernetically all three constituent concepts can be defined in
terms of cybernetical control, where attachment is a feedback-controlled system and exploration
is a feedforward-controlled system. The law of Requisite Variety, which is the concept
underpinning the vertical axis, states that “if a system is to be stable the number of states of its
control mechanism must be greater than or equal to the number of states in the system being
controlled” (Ashby, 1956). This means for the human mind, as an emergent property of the
dynamic interactions of the canonical structure of feedforward and controlling feedback loops in
the brain ((Shepard, 2004) that the more complexity it can be part of, the more complexity it will
be able to navigate. This vertical axis is designed for measuring how people deal with the
complexity in their environment, or, in other words, the ecosystem they are living in and are part
of. Those, scoring “low” against the background of the normgroup, are likely to be somewhat less
integrated with their social ecosystem, which allows them “positively” to focus more on unique
contributions to the system like being very creative, but might also “negatively” be perceived as
self-centered.
The AEM-cube® was designed based upon the observation that all the processes creating
the human mind as well as all the processes creating complex dynamic systems - like man-made
organizations or natural eco-systems - can be described as being defined by combinations of
feedback- and feedforward loops. These combinations are creating and can be described as
emergent processes and, if stable, as attractors (Robertson, 2005).
The self-organizing time-dependent shift from feedforward-control to feedback-control can
be linked to the Growth-curve or S-curve (see for further explanation below figure 2). These
Growth-curves have characteristics in common, whether observed in nature, organizations,
markets, products or services. This observation made it obvious to create a single frame of
Copyright © 2013 P. Robertson
Page 5 of 30
How People Contribute to Growth-curves
reference for seemingly so different systems as the human mind and organizational phases of
growth (Robertson, 2012).
Feedforward-control can also be replaced by information-control and feedback-control by
error-control (Pribram, 1976). This makes it easy to connect more information driven topics like
vision, mission and strategy to feedforward-control and more error driven topics like financial or
security management to error-control.
These direct links between human personality and time driven processes - like the Growthcurve - have been translated in a practical way to a three dimensional model called the AEMcube® (Robertson, 1999, 2003, 2005) The more cybernetical and scholarly phrasing above can be
translated in a more practical way, appealing to daily use in the executive and organizational lane.
In the Figures 1, 2 and 3 below, the three key questions connecting the human mind with
organizational processes are described in the language that evolved in the daily organizational
practice. The three questions defining the AEM-cube®’s relationship with the Growth-curve are:
1. Where do people contribute optimally to the Growth-curve?
2. Is the contribution focused on relationships or content?
3. Is the contribution integrating of differentiating?
The AEM-cube® is administered via a web based questionnaire, on average within 10
minutes. In most cases, respondents fill out a self-perception and also receive a (combined)
feedback-profile from co-workers or other people in their environment. As such, the AEM-cube®
makes use of the so called 360°methodology (Conway & Huffcutt, 1997).
Three main questions
The first question the AEM-cube® is likely to provide an answer for - “Where do people
contribute optimally to the Growth-curve?” - is scored by the right-left axis of the AEM-cube®.
The position on the plane of the AEM-cube® mirrors the contribution to the S-curve position. In
other words, a score to the right (the exploratory side) is related to feedforward-steering
characteristics of a personality and is in sync with feedforward-steering characteristics as they are
typical for the early stages of a growth-curve. The more a respondent scores on the left hand side
of the AEM-cube®, the more this is related to feedback-controlling characteristics of a personality
and in sync with feedback-controlling characteristics as they are typical for the later stages of a
growth-curve.
Copyright © 2013 P. Robertson
Page 6 of 30
How People Contribute to Growth-curves
Figure 2 Optimal contributions to the Growth-curve
The link between the growth-curve and the contribution of the human mind to the growthcurve is an essential and unique feature of the AEM-cube® approach. It is based upon shared
cybernetical foundations.
The cybernetical foundations of the human mind can be probed from the
neurophysiological (brain) as well the ethological perspective. The cybernetical foundations of
the Growth-curve can be probed directly from cybernetical considerations.
With regards to the neurophysiological aspects of the brain, the fundamental unit of
operation in the brain is itself a dynamic feedforward- feedback system. These units are “the
elementary” building blocks of all processes in the brain (Shepard, 2004). The mind, being an
emergent property of the dynamic networks (Sporns, 2010) created by these cybernetical units, is
then overall controlled by several levels of cybernetical control as is in general the case in
complex dynamic cybernetical systems (Brooks, 1999). In other words, this emergent mind is
itself an cybernetical system which has been already postulated and developed since world war II
by for example Ashby, Walter, Wiener and Brooks (Pickering, 2010).
With regards to the ethological aspects of behavior it is, amongst others, Lorenz’ Nobel
prize winning work on ethology that deeply researched the nature of the exploration system in
animals and humans (Archer & Birke, 1983; Lorenz, 1981). Two citations from Lorenz, chapter 6
((Lorenz, 1981). Lorenz connects exploratory behavior to information processing, which is
Copyright © 2013 P. Robertson
Page 7 of 30
How People Contribute to Growth-curves
basically the same as feedforward-control (Pribram, 1976). Herewith to citations of Lorenz
(Lorenz, 1981):
“Being independent of any of the “common” motivations, exploratory behavior acquires a
kind of information that is in exactly the same sense objective as are the results of human
scientific results”.
“By responding to each single unknown object as if it were biologically relevant, these
animals unavoidable discover those things which really are relevant. This endows them with the
ability to adapt, through individual learning, to the most variegated biotopes”. Both Lorenz and
Archer point with their research to a strong relationship between information- or feed-forward
control and this exploratory human instinct. Whereby Archer points out how fundamentally
different the exploratory instinct is, compared to other human instincts (Archer & Birke, 1983).
The cybernetical underpinning of Growth-curves is already understood for decades
(Kefalas, 1978). This includes the dynamic gradual change from positive feedback (creating the
first inflection point) to negative feedback (creating the second inflection point).
If there is only feed-forward behavior a system reacts in a pre-defined way without
responding to what the effect will be. Although feed-forward behavior can happen throughout
most of the Growth-curve, it is obvious to be strongest at the beginning and the early phases
because there is already something going into a direction, but with almost no feed-back, given the
fact that feed-back can only exert influence if there is already something to modulate and if there
is already a structure to do that very modulation.
In a feed-forward system, the control variable adjustment is not error-based. Instead it is
based on information about the process, especially where it is going. Feed-back control is errorbased and feed-forward control is information based (Pribram, 1976)
The cybernetical base underpinning the Growth-curve, makes it logical that during growth,
which is a process of adding more error-control modulating and stabilizing structure, it starts first
with a process tilted towards feed-forward (information) control and ends with a process tilted
towards feed-back (error) control.
It seems that most human beings and social animals do have a clear operating exploratory
system in their early youth. The expression of this exploratory system differs between human
beings in adult life. This is nature-based defined and nurture-based modulated. Some
personalities remain strongly feedforward-controlled, i.e. they remain being very exploratory.
Copyright © 2013 P. Robertson
Page 8 of 30
How People Contribute to Growth-curves
Depending on a variety of systemic factors (for example family system, education, social context),
human beings adapt and develop in their personalities cybernetical characteristics between
feedforward- and feedback-control. These characteristics are very stable as will be shown later in
this article.
Regarding the AEM-cube® Growth-curve relationship along the exploratory-stability axis it
can be summarized that the cybernetical make-up of both the brain, Growth-curve and the human
instinctive exploratory system is the foundation of this first axis of the AEM-cube®.
The second question the AEM-cube® is likely to provide an answer for is: “Is the
contribution of people focused on relationships or content?”
It is not that difficult to recognize two main patterns of growth-curves in organizations: one
is the unique competency (Hamel & Prahalad, 1990), which is technical, or expertise based, in
short, content based, and the other is customer, client, user based, in short, relationship based.
This practical division between content and relationship based processes is based upon the
human attachment system (Bowlby, 1969; Lorenz, 1981). This is an instinctive feedbackcontrolled system (similar to the food, temperature, blood pressure or sexual system), that
searches for proximity with patterns it has defined, early in life, as familiar. These familiar
patterns serve as a set point defining what is still familiar and what is not. If an individual
encounters unfamiliarity or unknown factors, this attachment-system will define by its structure
and experience whether an individual starts searching for the proximity of familiar patterns.
These patterns, individuals are attached to, are often other human beings (parents, family,
friends). Lorenz (ibid) discovered that these patterns do not necessarily have to be the parents or
other people, but can basically consist of any pattern, as long as enough time has passed to
become defined as a familiar pattern early in life. Lorenz showed that animals can become
attached, and can be made to attach, to other animals or even non-living objects. Bowlby (1969,
1973, 1980) also stated that full attachment towards non-human beings is possible.
In the late eighties and early nineties - at the National Aerospace laboratory of the
Netherlands (NLR) – observations were made during coaching, counseling, interviews and
psychotherapy. In some cases, mental depression could, for example, be connected to having lost
the access to a software platform. This made a convincing, be it a quite practical, case for
applying the concept of matter-attachment to human beings (personal communication).
Copyright © 2013 P. Robertson
Page 9 of 30
How People Contribute to Growth-curves
In the AEM-cube® this resulted in a scale ranging from “matter-attachment to peopleattachment” to define characteristics of the attachment system of individuals.
For all practical purposes, the matter-attached versus people-attached continuum has been
translated in a preference for content versus a preference for people. The latter words give in the
daily use of the tool an easier way to quickly understand the applicability of the concept.
(Robertson, 2005).
Scoring at the front of the AEM-cube® is related to having more matter-attached
characteristics and scoring at the back of the AEM-cube® is related to having more peopleattached characteristics. This is illustrated in Figure 2 above, where the odd numbers 1, 3, 5 and 7
are typically matter-attached positions and the even numbers 2, 4, 6, 8 are typically peopleattached positions.
The third question the AEM-cube® is likely to provide an answer for is worded as: Is the
contribution of someone integrating or differentiating? In other words, what is the degree of
complexity-maturity in an individual? Asbhy’s Law of Requisite Variety (Ashby, 1957) is used
as a base of assessing the capacity to steer into complex environments.
In the use of the tool, words have been chosen that focus on an individual perspective
versus a group perspective. This translates in a concept that runs from differentiating (= low
complexity-maturity) to integrating (= high complexity-maturity).
People who score low on this vertical axis are less focused on the whole ecosystem and
more focused on their own contribution and people who score high on this vertical axis are highly
focused on the whole of the organizational ecosystem around them and the integration of their
own contribution where that matters.
An easy way of explaining this vertical axis, as reported by users, is to mention that the
total length of the vertical axis can be imagined as the length of a whole S-curve. If one scores
high on the vertical axis then one can connect with all people who contribute everywhere on the
S-curve and as such communicate with ‘the whole Growth-curve in action’. If one scores low on
the vertical axis, one is focused on her/his specific contribution on that part of the Growth-curve.
In Figure 3 below, the positions 1 and 3 are showing a short vertical axis, which is for all
practical purposes related to a more ‘in depth’ focus on a specific phase of the Growth-curve.
Position 1 is related to positions at the start of a Growth-curve. Position 1 could be related
to a specialist inventor or expert strategic analyst. Position 3 could be related to a specialist
Copyright © 2013 P. Robertson
Page 10 of 30
How People Contribute to Growth-curves
accountancy or assurance professional. Position 1 and 3 are then characterized by a focus on the
job and less by integrating the job into the whole Growth-curve. This is often positively
associated with a high level of expertise.
Figure 3 The vertical axis - Complexity-Maturity - of the AEM-cube®
Position 2 is an example of a position that is capable of integration of all people
contributing along the Growth-curve. This is a more generalist approach, connecting and
integrating, and far less a specialist approach.
Whilst (as will be illustrated below) the position on the bottom plane both on the
exploratory-stability axis (right-left) as well on the matter-people attachment-axis (front-back) is
stable during lifetime, this vertical axis can change during lifetime. There is, however, not a
necessary, mandatory or natural change from low complexity-maturity to high complexitymaturity. Although high complexity-maturity is important for aspects like overall performance,
leadership, conflict resolution, the low complexity-maturity is essential for creativity, individual
in depth expertise and high skilled specialisms and crafts. There need to be diversity in a team on
this axis like there is a need for diversity on both other axes.
After this brief overview of the key parameters measured by the AEM-cube® we can
illustrate the essence of Thomson’s work with the continuing of the early founding context of
Yahoo.
In Figure 4, the whole group of the early days is shown again in the full three dimensional
format. From this figure it becomes clear that Yang and Filo are not only working in an
exploratory way at the beginning of the Growth-curve, but also that they show up as scoring
amongst the highest percentiles on matter-attachment and low on the vertical axis, related to their
Copyright © 2013 P. Robertson
Page 11 of 30
How People Contribute to Growth-curves
specialist focus. This is a very typical position for people creating realistic exploratory and
recognized innovative discoveries ready to become transformed in commercial value propositions.
So far, no observations have been from the AEM-cube® database made during the last 17 years of
high tech inventions not coming out of this right-front low vertical position.
The two profiles, are epitomizing the research done by Thomson (ibid). They show Koogle
and Mallet contributing to the early and later stages of the growth-curve, whilst Koogle shows up
more on the exploratory commercial focused relationship side and Mallet more on the content
focused stability (procedural, quality, governance focused) side of the AEM-cube®.
Figure 4 Classical Founding Team Dynamics (Example: Yahoo)
Both Mallet, as well as Koogle, obtained high scores on the vertical axis, which reflects
their connecting attitude creating a dialogue between the inside and the outside focus of the
organization.
Since this research was published, Yahoo stalled its growth (although it remained a strong
firm). A hypothesis is that, since Yang became CEO, the focus might have been too much on
technological exploration and less on turning innovation into real commercial value propositions.
For the latter it is likely that organizations need the dialogue embracing the whole Growth-curve
and not a specialist focus on one part of it.
A more recent example will illustrate this point even more clear. AEM-cube® profiles of
Steve Jobs and Tim Cook were obtained at the time of the handover of the stewardship for Apple.
A few people - outside Apple – scored the questionnaire with no more information than available
for the public eye. Their profiles were confirmed by a group of people who are proficient with the
Copyright © 2013 P. Robertson
Page 12 of 30
How People Contribute to Growth-curves
AEM-cube®, but not closer to Apple. Although this example is mainly illustrative, outdated and
not scientific (see disclaimer) it has shown during many lectures to be useful in explaining the
1
Innovation
2
Optimisation
dynamics of the AEM-cube®.
2
Jobs
Cook
1
Figure 5 A highly functioning dialogue at Apple
Both positions should not surprise. The perception of Steve Jobs is, as in the Yahoo case,
on the familiar position where most of the real technological discoveries and innovations come
from. The perception of Tim Cook is on a familiar position for an operational role. Both the
perception of Steve Jobs as well as the perception of Tim Cook are scored strongly matterattached, which should not surprise either. It seems to be a key condition to really develop a
technological passion. Most important though is that they are observed to be connecting and in a
continuous dialogue spanning the early and later stages of the Growth-curve aligning the
performance from idea to asset.
Given the fact that now Tim Cook has the stewardship of Apple, the question arises what
such a shift in leadership would do for the organization. Despite the fact that operational
activities, customer, political and financial stakeholder relationships are appreciated better,
governance is done with more stability, and that given the pipeline there is a proper execution of
bringing good ideas with a strong brand to a healthy and willing market, it might be questioned
Copyright © 2013 P. Robertson
Page 13 of 30
How People Contribute to Growth-curves
whether this will maintain the real innovative power of Apple. Tim Cook is simply not that
person (Kelly, 2012). So for the short term, i.e. a couple of years, his focus on the second half of
the growth-curve might be very fruitful for shareholders in the first years of his remit, but they
should expect probably a lower level of innovation if Cook is not capable of keeping the
innovation going. It is probably a priority, but it does not seem to be his nature (Kelly, 2012). It
seems appropriate to state here already a point that will be worked out in the applications
mentioned in the conclusion, that someones position on the Growth-curve is not a judgement
about someones capacity to be a CEO. The message is, for any CEO, to align the contributions to
the Growth-curve with personalities that create together a working and sustainable growth path.
The examples show a relationship between personality and the Growth-curve. Since the
start of the development of the AEM-cube®, it has become evident that this relationship has a
practical implication for connecting human resources directly to strategic growth-phases that
exist in organizations.
A key question however is whether these personality patterns are stable or not. Simply
stated: is a profile, like that of the perception of Tim Cook, capable of change into a more
visionary, strategic contribution to the Growth-curve, or is his operational strength the talent he
brought, brings and will bring into the future?
The common sense is that these characteristics are stable: “an inventor will never become
an accountant and an accountant will never become an inventor.” The difference between these
two personalities is basically their cybernetic make-up.
Theoretically, the characteristics of the attachment pattern should be most stable, being
formed in life quite early (Bowlby, 1969). For the exploratory system this is less evident. Most
human beings express some level of exploratory (feed-forward-steering) behavior and it seems
that depending on the factors related to education, upbringing, family system, some express these
characteristics all their life and others show more stability focused (feed-back-controlling)
behavior. Although, since the almost two decades the AEM-cube® is in use, it is common sense
and daily practice to consider these characteristics to be stable.
Since, as of today, based upon more than ten years of data gathering, it is now possible to
get a significant number of assessments and re-assessments, to check this hypothesis about the
stability of the personality patterns created by the attachment and exploratory instinctive systems.
Copyright © 2013 P. Robertson
Page 14 of 30
How People Contribute to Growth-curves
The common sense is also that the vertical axis can be changed by personal choices and
development. Although a low vertical axis is productive from a focus on specific competencies
and specific in depth contributions to a Growth-curve, it is likely that people once they get
broader and larger responsibilities should grow to a higher level of complexity-maturity and
changing from a more specialist to a more generalist approach.
Based upon the considerations above a key question to be researched is, whether the bottom
plane of the AEM-cube® shows stable characteristics over time.
Statistical properties
Before approaching the question about the stability of the AEM-cube® profiles over time,
this article will first introduce a brief introduction about the statistical basics of the AEM-cube®.
Several data sets were statistically analyzed, where the data stem from employees from
many companies from mostly Anglo-Saxon countries. By combining several data sets with AEMcube® data at the item level, a final sample was reached containing 7.983 Self images and 21.605
Feedback images (in the text also called ‘Other’ assessments).
General statistics
In the following tables, some statistical and psychometrical properties of the tool are
summarized. The current version of the AEM-cube® sports 12 items for the A- and E-scale and
24 items for the Complexity-Maturity scale.
Self (N = 7.983)
Attachment
Exploration
Commaturity
Average
52.4
30.0
115.2
Standard deviation
8.8
9.4
12.2
Skewness
-0.4
0.6
-0.3
Kurtosis
0.1
0.3
0.3
Minimum
15
12
60
Maximum
72
68
144
Copyright © 2013 P. Robertson
Page 15 of 30
How People Contribute to Growth-curves
Other (N = 21.605)
Attachment
Exploration
Commaturity
Average
48.8
34.1
109.0
Standard deviation
10.8
11.8
17.4
Skewness
-0.4
0.5
-0.6
Kurtosis
-0.2
-0.2
0.6
Minimum
12
12
24
Maximum
72
72
144
Attachment
Exploration
Commaturity
Average
49.8
33.0
110.7
Standard deviation
10.4
11.4
16.4
Skewness
-0.4
0.6
-0.7
Kurtosis
-0.1
0.0
0.9
Minimum
12
12
24
Maximum
72
72
144
All (N = 29.588)
A
E
M
Table 1 AEM-cube®: Moments of the scales in full dataset in three groups
The most prominent feature of all three scales is that almost the full range of possible
scores per scale is visible (12 to 72 for the A- and E-scale and 24 to 144 for the M-scale). This
means that respondents are able to make clear distinctions when describing their own behavior or
the behavior of others. So, statistically speaking there is variability (‘variance’) which is a first
prerequisite of any measurement tool.
When looking at the third and fourth moments (skewness and kurtosis) of the scales in the
three groups, no strong deviations from a standard-normal distribution are observed (all values
are in the -1.0 to +1.0 range). This is statistically attractive as many psychometrical or statistical
methods assume such a standard-normal distribution. Although the Kolmogorov-Smirnov test
statistic is significant for all three scales, this is due to the large sample size. These statistics are
about 0.07. To see a graphical indication of the normality, see the next Figure. Herein the
expected versus observed data for the first scale (Attachment) are plotted. As one can observe,
the deviations between Expected and Observed scores are quite small.
Copyright © 2013 P. Robertson
Page 16 of 30
How People Contribute to Growth-curves
Figure 6 Expected versus observed data for Attachment scale (example)
Reliability of the scales
The next step in the analysis is to take a look at the reliabilities of the scales. Reliability is
defined as ‘internal consistency’, meaning the degree the items that make up the scale indeed tap
a common construct (Cronbach, 1951; Hofstee, 1966). In statistical terms one would expect the
items to correlate. Although there exist several methods to compute ‘the’ reliability of a scale,
like splitting the items in two groups and comparing the total scores on these halves, the best
measure here is Cronbach’s alpha. This statistic tells us what the average correlation between all
possible split halves of the items would have been. So, in most psychometrical texts this statistic
is used. In Table 2 this statistic is shown for all three groups.
Group
Attachment
Exploration
Commaturity
Self
0.88
0.92
0.88
7.983
Other
0.93
0.94
0.93
21.605
All
0.92
0.94
0.93
29.588
12
24
Items
12
N
Table 2 AEM-cube®: Reliabilities of the scales in three groups
The table shows very strong alphas. In fact, the alphas demonstrate that the items of each
scale are very closely related. One could argue that the alphas are too high, meaning that 'the
Copyright © 2013 P. Robertson
Page 17 of 30
How People Contribute to Growth-curves
same question is asked over and over again'. This leaves room for a considerable reduction of the
number of items per scale.
The next step in test construction is to take a look at the latent structure underlying the
instrument. The techniques most often used are called factor analysis, which comes in many
different flavors (Thompson, 2004). The variant mostly used is principal components. In the table
below the outcome of such principal components analysis is shown (N = 29.588). The component
extraction was forced to three factors, followed by a (orthogonal) Varimax rotation. The
following table shows a very strong latent structure, which is conforming to the expectations. The
table has been split and both halves are presented next to each other for layout reasons. The
current representation is in a fact a condensation of a table that is in fact twice as long. Loadings
smaller than | 0.40 | are omitted.
In 2006 a (smaller) data set was split into four subgroups (Self and Feedback plus odd or
even row numbers of the respondents). On these four subgroups the same principal component
analysis was carried out, separately. It turned out that the solutions found in all groups were
similar. So, the results as shown in Table 3 were cross-validated. For more details see
Schoonman (2006).
Copyright © 2013 P. Robertson
Page 18 of 30
How People Contribute to Growth-curves
Item#
M
E
A
Item#
M
01
0.40
25
0.58
02
0.62
26
0.52
03
0.48
27
0.53
04
0.55
28
0.54
05
0.58
29
0.63
06
0.60
30
0.41
E
07
0.79
31
0.76
08
0.78
32
0.75
09
0.76
33
0.79
10
0.78
34
0.65
11
0.80
35
0.74
12
0.73
36
0.77
13
0.62
37
0.66
14
0.68
38
0.67
15
0.53
39
0.60
16
0.49
40
0.65
17
0.62
41
0.64
18
0.67
42
0.66
A
19
0.76
42
0.73
20
0.74
44
0.60
21
0.64
45
0.78
22
0.61
46
0.75
23
0.70
47
0.67
24
0.81
48
0.62
Table 3 AEM-cube®: Varimax rotated solution with three forced components
In the ideal case, the researcher wants to see that all items from the intended scale indeed
have the highest loadings on one and the same component and smaller or no loadings on the other
ones. In the table one observes exactly this. The results show clearly that a latent structure is
present and can easily be interpreted.
Test-retest stability of the scales
Copyright © 2013 P. Robertson
Page 19 of 30
How People Contribute to Growth-curves
A subset of respondents filled in the AEM-cube® twice (N = 98). The interval between the
two (Self) images varied between 1 and 12 years, with an average interval of 2.5 years. In the
next table the correlations between the two scores obtained on the two administrations is shown.
Also, the average absolute score difference in percentiles is shown. The last rows shows the
percentages of respondents in three different score groups. These score groups are calculated as:
low (absolute score difference in percentiles between 0 – 10)
med (absolute difference between 11 – 20)
high (absolute difference 20+)
Retest group (N = 98)
Attachment
Stability
0.83
Exploration
Commaturity
0.81
0.73
Absolute score difference
10
11
13
Low difference
75
61
56
Med difference
12
25
16
High difference
13
13
27
Table 4 AEM-cube®: Stability in time (retest reliability)
This table makes clear that the obtained scores are stable over time. Between 72 to 87% of
the respondents will obtain a score at the second administration that is differing less than 20
percentile points compared to the first administration. The correlation coefficients tell the same
story. The scores are rather stable in time. Complexity-Maturity has the lowest stability, which
was the expectation, because, based upon the law of Requisite Variety, it can be influenced and
increased by learning. In order to assess these changes more in depth research is under way to
connect the effect of leadership programs to changes on this scale (Pinckers, Vebego,
Netherlands to be published 2014).
Discriminant validity
The next step is to check the discriminant validity, e.g. the degree in which the scales are
invariant or uncorrelated to each other (Campbell & Fiske, 1959). This is a desirable property of
any multi-scale instrument. Discriminant validity is important as one does not want to have
independent measures to be correlated. The table below shows the intercorrelations between the
Copyright © 2013 P. Robertson
Page 20 of 30
How People Contribute to Growth-curves
three scales, in both versions (Self ratings above the diagonal (N = 7.983), Feedback ratings
below; N = 21.605).
Discriminant
Attachment
Attachment
Exploration
Commaturity
Exploration
Commaturity
-0.21
0.33
-0.26
0.49
-0.39
-0.42
Table 5 AEM-cube®: Correlations between scales (discriminant validity)
The scales have between 4 and 25% of variance in common. As the problem mainly lies in
the correlations between the A- and E-scale with the M-scale, one could consider reducing the
number of items for the latter. This could be done by including in the shortened scale only those
items that have low correlations with the total scores on the A- and E-scale. A positive side-effect
would be the reduction of items in the M-scale which has now 24 items, whereas the A- and Escale have 12. It was shown elsewhere that the number of items could be reduced to 3 x 8 = 24 in
total, without a loss of reliability (Schoonman, 2013a).
Self versus Other ratings
When comparing scores of Self-images with Feedback-images by means of the Students ttest (Ferguson, 1976), the following table arises:
Self - Other
t-value
Attachment
Exploration
Commaturity
Degrees of freedom
Significance
Score difference
29.9
17306
.00
3.7
-31.5
17670
.00
-4.2
29.2
20149
.00
6.2
Table 6 AEM-cube®: t-test of differences between Self and Feedback ratings
Although the t-test is rather sensitive for type I errors (incorrect rejection of the null
hypothesis) when samples become larger, significant differences are found when people rate
themselves or when people rate others. In general, Self-ratings are higher on Attachment and
Complexity-Maturity (some 5 points on each scale), whereas Self ratings on Exploration are
Copyright © 2013 P. Robertson
Page 21 of 30
How People Contribute to Growth-curves
lower. However, if effect sizes (what is the practical meaning of significant differences) are
calculated, this results in the following D-statistics (Cohen, 1992):
Attachment: D = 0.16
Exploration: D = - 0.17
Maturity: D = 0.17
These D-values fall in between the effect sizes ‘small’ and ‘medium’, so a certain effect of
who is responding to the questions is visible. This justifies the use of the 360º methodology: in
case no differences would be visible one source (rater) would suffice. To make the scores from
different sources comparable, different norm groups (comparison groups) are used for both types
of images in the instrument. More on norming issues can be found in Schoonman (2013b), page
50 and further.
Gender differences
From a small sample (N=199) done in the early stages of development at the Erasmus
University of Rotterdam (Olde Bijvanck, 1997) the following outcomes regarding gender
differences were expected:
Women are slightly less exploratory than men
Women are more people-attached than men
Women score about equal on the complexity-maturity scale compared to men
The table below shows the results of the t-test, computed over the total sample (Male N =
19.042; Female N = 10.546). Again, the differences are significant, but, when looking at the
average score difference (last column in Table 7) the differences are of no practical relevance.
The effect sizes (Cohen, 1992) are minimal.
Gender
t-value
Degrees of freedom
Significance
Score difference
Attachment
-21.4
21035
.00
-2.7
Exploration
5.2
21176
.00
0.7
-9.2
21118
.00
-1.8
Commaturity
Table 7 AEM-cube®: t-test of gender differences
Copyright © 2013 P. Robertson
Page 22 of 30
How People Contribute to Growth-curves
The sample group consists mostly from working people: self employed, or at small
businesses and large corporations. Hardly unemployed or housewives/men make part of the
sample.
Male
Female
A achment
Explora on
Maturity
Figure 7 Gender differences in total sample
The data suggest that there is no real difference between male and female respondents. But
the differences, although small, look aligned with the expectations.
This holds true at least for the norm group population where the male/ female ratio is 2:1.
As mentioned above, we are observing here a working population.
Educational differences
Some psychometrical tools are sensitive to educational level of the respondents, or
indirectly with cognitive abilities (Jensen, 1980). Often this is not desirable, as a measurement of
one aspect should be as invariant as possible from other individual characteristics. For this reason,
the next analysis was carried out. From the majority of the respondents the educational
background level is known (N = 3.672 = 12% missing at a total of 29.588). The educational
backgrounds are divided into seven levels. These levels were condensed into three groups: low,
medium and high. The next table shows the average scores on the scales in these groups.
Copyright © 2013 P. Robertson
Page 23 of 30
How People Contribute to Growth-curves
Education
Low
Medium
High
Attachment
49.9
50.1
49.7
Exploration
34.0
32.9
32.6
Commaturity
110.5
111.4
110.6
N
3018
10814
12084
Table 8 AEM-cube®: Average scores in three educational levels
It is clear that these differences are negligible, so the conclusion can be drawn that the
instrument is invariant with regard to educational level of the respondent.
Summary of statistical basics
To summarize the statistical analyses carried out thus far:
The internal structure of the instrument, as shown by distribution of scores, internal
consistency (reliability) and latent structure is well within psychometrical standards.
The discriminant validity of the AEM-cube® is satisfactory but could be further improved by
reducing the number of (certain) items of the M-scale. This would reduce the common
variance with the other two scales and would, at the same time, increase the practical value
of the instrument.
The stability of the scores in time is a remarkable property of the instrument. In a time
interval of about 2.5 years, scores on the three scales do not differ more than 20 percentiles
for the majority of respondents. This might show that the concepts being measured indeed
have an ecological (or biological) foundation, founded in early life.
The instrument looks invariant with regard to gender and educational level. This is an
attractive feature as no separate norming with its problems have to be applied. With regard to
differences between Self and Feedback assessments, small to medium effects were found.
This implies that the current separate norming should stay in place.
Conclusion
There is now increasing evidence that it is possible to construct a valid tool, like the AEMcube®, that is capable to describe human personalities based upon a falsifiable cybernetical
foundation. In other words: three non-psychological concepts are able to describe a coherent set
of personality characteristics. The use of the concepts in this combination are unique in the field
Copyright © 2013 P. Robertson
Page 24 of 30
How People Contribute to Growth-curves
of (organizational)-ecology (where they probably belong best) and even more so in the field of
psychology.
A practical consequence from this cybernetical approach is that personality can be
described in a frame of reference that might be aligned to basically any complex living system, or
in other words any ecosystem. Hence, for example, the connection with the Growth-curve.
The statistical foundation of the model is aligned with general accepted norms. The
practical applications seem to be a spin-off from this solid base in a rather non-linear fashion. The
link with the Growth-curve, which surfaces since the last decade as one of the most practical
spin-off concepts is falsifiable based upon the fact that both Growth-curves as well human
personalities can be described as a dynamic mix on a continuum from feedforward-controlled
characteristics to feedback-controlled characteristics. The ‘popularity’ of this concept in daily
practice, and the way the Growth-curve concepts connects the concepts together was rather
unforeseen, but in hindsight, as so often, obvious.
In order to support the common-sense and fast growing anecdotic evidence (Robertson,
2005; Thomson, 2006) it will be necessary to aggregate the existing and future material into a
batch of data for further analysis (see also the discussion below).
Most standing out is the result that over time these cybernetical characteristics of human
personalities do not change significantly over time.
For this article the focus is mostly on the ‘bottom-plane’ of the AEM-cube® and it has
become clear that the positions are over time very stable. This was expected, based upon the
foundational ethological concepts and the coaching and counseling observations of thousands of
managers.
Practical applications of the AEM-cube® can be observed now developing themselves in
the following fields:
-
Career and Talent Coaching. If it is known to what phase of any Growth-curve someone
contributes to, it is likely best practice to coach people towards that optimal
contribution in life. Research is under way with a cohort of MBA students to investigate
the usefulness of this approach as support of their career strategies (Miller & Robertson,
MIIS, to be published). Other individual approaches in the field of counseling and
mentoring are explored in the USA, UK and the Netherlands by small cohorts of
experts.
Copyright © 2013 P. Robertson
Page 25 of 30
How People Contribute to Growth-curves
-
Investor-Start up cooperation. Based upon the start-up research in the USA (Thomson,
2006) and case by case studies in the Netherlands and the UK patterns of the most
effective alignment of team members along the Growth-curve are being discovered.
This is one of the promising fields much research will be focused on (see discussion
below)
-
(Top) Management Teams. Management teams have in general to deal with a strategic
context that can be framed in terms of the always moving Growth-curve. Matching the
team alignment toward the ever changing strategic context seems to be one of the most
obvious applications of the AEM-cube® approach. This does not mean automatically
changing people. If a team is, in terms of the AEM-cube® divers enough, there should
often be enough resources to cope with a changing environment.
-
Strategic Human Resources Management. There are now several companies exploring
and improving the AEM-cube® profile as a standard dataset in their HR database in
order to create an optimal strategic team alignment whilst creating teams or changing
teams by succession. In this situation, if people need to change assignments or jobs, an
AEM-cube® profile history is already available.
-
A critical note is to be made about using the AEM-cube® for external recruitment. It is
attractive to mention this option. Research has been done in the Netherlands (GITP)
showing the AEM-cube® method as solid as other methods, but like any other approach,
it should only be used in professional settings and as part of a multi-level assessment
where more tools, assessments and interviews are used. With the limits guarding this
professional and ethical background the AEM-cube® is used already for more than five
years in recruitment at GITP, Netherlands.
It is likely that the AEM-cube® has the potential to contribute to a long term strategic
human resources management both from the organization to optimize individuals contributions to
the strategic phases of growth, as well from individuals to organizations to optimize their
individual career paths and their individual contributions to growth.
Discussion
From a psychometrical point-of-view, the statistics calculated on a very large sample and
presented in this article might be called impressive and the longitudinal stability outstanding. In a
Copyright © 2013 P. Robertson
Page 26 of 30
How People Contribute to Growth-curves
statistical way, the instrument at hand has a solid basis for application in ‘real life’ situations. The
internal structure is up to standards (Bartram, 2004; Lindley, Bartram, & Kennedy, 2008). Also,
the theoretical foundation seems to based on strong concepts, fitting in existing ‘nomological
networks’.
Based upon this strong foundation, the priority of future research must now be focused on
strengthening the construct and predictive validity with more empirical evidence.
Evidence can be found in the relationships found with other instruments (Schoonman, 2007,
2008), and more practical evidence in the research into High Tech Start-Ups (Thomson, 2006).
The claims regarding the correspondence between the exploratory-stability scale and Growthcurves are based upon a falsifiable cybernetical concept. They are supported on a case by case
base and the consolidation of this case material is a logical next step. Further it will be a priority
to follow-up with specific research in ventures and organizations, needed to create a further
strengthening of concept and a reference base for the many possible practical applications.
Correspondence:
Peter P. Robertson: probertson@hi-int.com
Acknowledgements:
Richard M. Robertson, MSc. Affiliation: Nyenrode Business University, Breukelen, The
Netherlands
Disclaimer
The business examples shared in this article are used for educational purposes. Factual data are
highlighted in the context of this purpose and the writers do not take responsibility for any other
use of these data. The Apple and Yahoo case are now outdated and lost their relevance beyond
the educational, which is the purpose of this article. The Yahoo case is unchanged since its
publication in Thomson's book (see reference in text). The picture was not published in the book,
but the text in that book is referring to this case.
Copyright © 2013 P. Robertson
Page 27 of 30
How People Contribute to Growth-curves
For Apple, the data are illustrative, but not created by people from within: observations were
outside-in and generated based upon what the public eye can see. The number of participants
showed a statistical consistent picture, solid enough to use for educational purposes, but by no
means useful for further interests. The authors do not take any responsibility for use of these
cases for other purposes than the educational purpose stated above.
Copyright © 2013 P. Robertson
Page 28 of 30
How People Contribute to Growth-curves
References
Archer, J., & Birke, L. (1983). Exploration in Animals and Humans. United Kingdom: Van
Nostrand Reinhold Co.
Ashby, W. R. (1956). An Introduction to Cybernetics (Internet (1999):
http://pcp.vub.ac.be/books/IntroCyb.pdf ed.). London: Chapman & Hall.
Ashby, W. R. (1957). An Introduction to Cybernetics. London: Chapman & Hall.
Bartram, D. (2004). EFPA Review model for the description and evaluation of psychological
tests (Vol. 1.0). Brussel: EFPA.
Bowlby, J. (1969). Attachment and Loss (Vol. 1. Attachment). New York: Basic Books.
Bowlby, J. (1973). Attachment and Loss (Vol. 2. Separation, Anxiety and Anger). New York:
Basic Books.
Bowlby, J. (1980). Attachment and Loss (Vol. 3. Loss). New York: Basic Books.
Brooks, R. A. (1999). Cambrian Intelligence. Cambridge: MIT Press.
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the
multitrait-multimethod matrix. Psychological Bulletin, 56, 2, 81-105.
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155-159.
Conway, J. M., & Huffcutt, A. I. (1997). Psychometric properties of multisource performance
ratings: a meta-analysis of subordinate, supervisor, peer, and self-ratings. Human
Performance, 10(4), 331-360.
Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika(16),
297-334.
De Wit, B., & Meyer, R. (2010). Strategy Synthesis. Resolving strategy paradoxes to create
competitive advantage. Andover: Cengace Learning.
Ferguson, G. A. (1976). Statistical analysis in psychology & education (4th ed.). Tokyo:
McGraw-Hill.
Hamel, G., & Prahalad, C. (1990). The core competence of the corporation. Harvard Educational
Review, 68(3), 79-93.
Hofstee, W. K. B. (1966). Models in split-half reliability. Nederlands Tijdschrift voor de
Psychologie, 21, 521-529.
Jensen, A. R. (1980). Bias in mental testing: New York: The Free Press.
Kefalas, A. G. (1978). Managing A Steady-State Firm. A Cybernetic Framework. Current Topics
in Cybernetics and Systems. In J. Rose (Ed.), Current Topics in Cybernetics and Systems
-Proceedings of the Fourth International Congress of Cybernetics & Systems 21–25
August, 1978 Amsterdam, The Netherlands (pp. 86). Berlin Heidelberg: Springer Verlag
Kelly, H. (2012). How Apple has changed under Tim Cook. Retrieved October 4, 2012, 2012,
from http://www.cnn.com/2012/10/04/tech/innovation/apple-tim-cook/index.html
Lindley, P., Bartram, D., & Kennedy, N. (2008). EFPA review model (update) (Vol. 3.42).
Brussel: EFPA.
Lorenz, K. (1981). The Foundations of Ethology. New York: Touchstone Books.
Modis, T. (1998). Conquering Uncertainty: Understanding Corporate Cycles and Positioning
Your Company to Survive the Changing Environment. New York: McGraw-Hill.
O'Reilly, C. A., Tushman, M.L. (2004). The ambidextrous organization. Harvard Business
Review, 82(4), 74-81.
Olde Bijvanck, D. (1997). Afstudeeropdracht: De AEM-cube ®. Rotterdam: Erasmus University.
Pickering, A. (2010). The Cybernetic Brain. Chicago: The University of Chicago Press.
Pribram, K. H., Gill, M.G. (1976). Freud's 'Project' Re-assessed. London: Hutchinson & Co.
Copyright © 2013 P. Robertson
Page 29 of 30
How People Contribute to Growth-curves
Robertson, P. P. (1999). The AEM-Cube; a management tool, based on ecological concepts, in
order to profit from diversity. Paper presented at the The 43rd meeting of the International
Society for the Systems Sciences, Asilomar Conference Center, Pacific Grove, California
USA.
Robertson, P. P. (2003). Ontsnappen uit S-catraz (Always Change a Winning Team). Schiedam:
Scriptum Management.
Robertson, P. P. (2005). Always change a winning team. Singapore: Marshall Cavendish.
Robertson, P. P. (2012). Why Top Executives Derail; A Performative-Extended Mind and a Law
of Optimal Emergence. Retrieved from http://dx.doi.org/10.2139/ssrn.2097240 website:
http://ssrn.com/abstract=2097240
Schoonman, W. (2006). The AEM Cube revisited. Some statistical exercises. Den Haag: Psy
Tech.
Schoonman, W. (2007). The AEM Cube constructs. Some notes on construct validity. Den Haag:
Psy Tech.
Schoonman, W. (2008). The AEM Cube constructs. In search of construct validity. Den Haag:
Psy Tech.
Schoonman, W. (2013a). Full Data Analysis of AEM Cube. Den Haag: Psy Tech.
Schoonman, W. (2013b). Mensen beoordelen. Voor HR-professionals (Vol. 2). Amsterdam: De
Witte Ridders.
Shepard, G. M. (2004). The Synaptic Organization of the Brain. New York: Oxford Unversity
Press, Inc.
Sporns, O. (2010). Networks of the Brain. Cambridge: The MIT Press.
Thompson, B. (2004). Exploratory and confirmatory factor analysis. Understanding concepts
and applications. Washington: American Psychological Association.
Thomson, D. G. (2006). Blueprint to a billion : 7 essentials to achieve exponential growth.
Hoboken, N.J.: John Wiley & Sons
Copyright © 2013 P. Robertson
Page 30 of 30