[5] Control Loops

For a large part, living means seeking goals.

At this and any other moment, sophisticated regulatory mechanisms in your organism are striving to keep physiological parameters, like body temperature, blood sugar level and countless other concentrations of chemicals, within a healthy range. Your organism is seeking the goal of healthiness, a state which is constantly threatened by many influences.

When you walk to the fridge for a drink (say, in order to satisfy an extra demand of water due to that salty pizza you ate before), a continuous stream of input signals from various types of sensory cells (eyes, equilibrium organ in the ear, ..) is processed by your nervous system and transformed into an output stream of muscle activating nerve signals. The activated muscles then create forces, change the body posture, and the effect is again monitored by the sensory cells. So there is a control loop from sensors to actors to the environment and back to the sensors. As a result of this control, your mechanically instable body does not tumble or hit any obstacles and simultaneously keeps approaching the fridge. This is goal-seeking, intelligent behavior - one of the big differences between living beings and inert matter.

All this happens automatically and normally does not require any conscious planing. But what about the more long-term goals in our life, such as reducing body weight, getting a project done, or becoming a happier person ? Why are such higher-level goals sometimes so hard to achieve ? Is there some fundamental difference between walking to the fridge and successfully finishing a complex long-term project ?

The question of how goals can be achieved, in a very general and abstract scope, has been considered in the scientific discipline of cybernetics, already many decades ago. Unfortunately, cybernetics did not survive as a coherent research field, but its ideas are alive and continue to evolve in different branches of modern science.

One of the most fundamental concepts of cybernetics is the feedback control loop. It is most often explained by the example of a temperature control systems of a house:

The feedback control system consists of a sensor, a goal, a processor, and actors. In our example, the temperature sensor measures the inside temperature of the house. The processor constantly compares this actual temperature with the goal temperature, which has to be set by the user in advance. If the actual temperature is lower than the goal value, the processor turns on a heating device (actor 1). If the temperature is too high, the processor activates a cooling device (actor 2).

This simple reactive control scheme has many great features. In particular, it does not require any understanding of the many possible reasons that may disturb the inside temperature of the house, such as outside weather changes, doors left open for too long, etc.. The feedback controller is monitoring only a single goal variable and applies one of only two actions whenever appropriate. In contrast, a control system based on prevention instead of reaction (also called feed-forward control) would have to take into account all those possible disturbances and provide selective counter measures for each of them.

Another advantage of feedback control is that it works even without any pre-arranged step-by-step plan of how to achieve the goal (Anyway, such detailed plans often fail in reality because there can occur completely unexpected problems in complex environments). Instead, it is often sufficient to re-evaluate the situation after each step, and to have at least one action available that brings one closer to the goal. Such an open strategy is much more robust than detailed planing.

Of course, feedback control also has its problems in practice. The corrective actions against some perturbation may come too late, or they may be in-appropriate in size: Both, too weak and too strong counter measures have to be avoided. It is therefore not surprising that optimum control theory is a rather complex field in applied mathematics.

Now, how can this concept of feedback control be applied to the case of the "higher-level" goals that we pursue in our lives ?

Let me start with a concrete example. During my 1.5 years long stay in California, I had developed overweight. After my return to Germany, I decided to reverse that trend by reducing the overall amount of food and by completely giving up a few particularly problematic habits (potato chips, etc.). Except these very few bad habits, there was no need to improve the quality of my daily food, since I was already a fan of healthy Asian dishes. I just got used to somewhat smaller portions. Also, I didn't do any sports (it only makes me hungry).

Over a little more than a year I successfully reduced my weight by 16 kg, and then I kept my goal weight for about 5 years. Then I lost control for a while and quickly gained 5 kg, but since the beginning of this year I applied the same fasting method and was able to bring my weight back to the goal value.

In retrospect, I realize that the success of this method was due to a few crucial design principles.

One success principle is the gradual effect of small persistent steps. I did not have to reduce the portions of my food by much, only by 1/3 or so. After a surprisingly short time I was getting used to the new standard amount of nourishment. I wondered why I ever needed to eat that much before. But the body still reacts, if the slight reduction is persistently maintained over a long period. I will write about the power of the gradual in another post (drop by drop).

But clearly, the main success principle was the establishment of a control loop:

A definite goal weight was set up in advance, the actual weight was monitored, and the available counter actions were a reduction or increase of the daily food portions.

This sounds trivial, just like ordinary fasting. But a key point for me was to take the monitoring of the actual weight very seriously. I bought a digital scale with 100g resolution and measured my weight EVERY morning at the same time, before the first sip of coffee. And I drew a graph of my weight as a function of time and looked at it at the beginning of each new day.

By looking regularly at this curve, you learn a lot about yourself.

You realize that there is a certain random component, which keeps fluctuating up and down, seemingly without correlation to your eating and drinking habits. It is important to get a feel for the size of these random fluctuations, in order not to be fooled by them.

But there is also a very clear deterministic component. With the time, you will see the effects of your behavior during the day before very clearly. Usually, even a slight breaking of your self-established rules will show up next morning. A persistent fasting period of a whole week can be destroyed (at least temporarily) by a single day of loosened control. This clear feedback, however, helps you to avoid such rule-breaking in the future, especially if you write down an explicit note that "behavior X has led to the undesired result Y" and read that note regularly.

On the other hand: What a pleasant and reassuring experience it is to observe a persistent downward trend in your weight curve over several weeks ! This is, after all, a direct, physical manifestation of sustained will power ! It shows that your method is actually working and that you can probably reach your goal by simply following this path further.

I strongly recommend to really draw such a progress curve, rather than just reading the list of weight numbers. The graphical representation is so much more impressive !

So, what we can learn from this successful implementation of a control loop is the importance of a clear monitoring of the effects of our actions. I think that applies for other long-term projects as well: You need to precisely measure and document your progress, or lack of progress, and it must be clear which actions (or missed actions) have caused which effects.

Unfortunately, this can be very difficult in more complex endeavors. Sometimes projects go wrong in a way that is hard to analyze and quantify.

For example, I repeatedly failed in bringing my research projects to a fruitful ending (meaning to write a publication). The reason was not that I couldn't find good solutions to the given problems, but rather that each solution was leading to a couple of new interesting questions. These new questions were often leading into a different - and much more exciting - direction, so that the original questions suddenly appeared quite inadequate and boring. With my mind already drawn into the new problem solving process, I completely forgot about the publication of the intermediate results. And so on and on.

Writing now about this silly behavioral pattern, it all appears obvious to me. But while in the actual process of research, one usually lacks the necessary distance and overview. One not even recognizes immediately that something is going wrong. One is completely absorbed in fascinating small problems.

In other words: There is no clear monitoring of overall progress.

In the present example, a proper goal would be a REGULAR publication of the results of one's scientific work. It is important to formulate the goal in a way that can be accomplished in practice. So, the goal is NOT to wait with the publication until the research projects will reach a self-contained state of perfect satisfaction, because one NEVER comes to a point where all aspects are 100% satisfying. It would be much better to impose oneself a strict time limit for the actual research work and then publish the present state of the project, no matter how many open questions remain.

Then, in the publication phase, the writing of the manuscript can be divided into smaller sub-tasks (e.g., individual sections or figures). One simply has to write a to-do list and finish point after point.

Again, the great danger here is that while writing the manuscript, one becomes increasingly aware of all the shortcomings of one's work. But now it is essential to remember the higher goal of regular publication and to resist the urge to improve the work by further research. Once a master plan has been thoroughly set up, one should stick to it.

As it often happens, also this little wisdom can be illustrated using the world of music. As a musician, one has only a limited time to learn and practice. This limited musical experience one brings onto the stage in a live performance. During the performance, one has to make the best out of these resources. It is useless to try to make up for shortcomings just a few hours before the show. AFTER the show there is time for new learning and improvement.

At the end of this post (I have set myself a strict time limit although the article is still far from perfect), I would like to invite the reader to figure out a control loop for the meta goal of becoming a happier person. How can we monitor the various aspects of our mental state over time ? Which of our actions are particularly relevant to change this mental state ? What are the perturbations like and what is your available set of counter measures ?