Why Models are Wrong and still Useful!

Reflections from the Q&A session at INSEE conference presentation on system dynamics and on occasion of completion of 1 year of system dynamics work at TERI.

IMG-20160106-WA0006_1

We always use models. Sometime we know that we are using them while most of the time we are not aware about it. Here is an example. When you go to buy something for a friend or family member how do you make a decision on what to buy? Suppose I go to a shop to buy a shirt for my father, I create a mental image of him in my mind. It is a model. I use it to see if the shirt will suit him. I am now creating simulations. Based on the results of my mental simulations I take a decision of either to buy the product or not. Is my mental model right? No. It is a reflection of reality. My image of my father is a partial reflection about him. But it is still useful. I cannot do away with it. Does that effect my decision? Yes. At times when I am not sure about how the shirt will look on my father, I ask the shop keeper if he would exchange it. On other days, I would shortlist the shirt and say I will come back with my father. What is happening here? I don’t have enough confidence on my mental model. But I am still using it to make a decision. It all depends on what model we have the highest confidence upon. We use them all the time. Even better, I would prefer to have my sister, mother or wife with me to make the decision of buying the shirt. I would consult them. I would want them to use their Mental Model and share their simulation findings with me. I am testing and validating my model. Does this make my model right? No. But it helps improve my understanding of my model. It helps me to make an informed decision. I use my mental model and other’s models to arrive at a conclusion. At times it works, at times it does not. My father may not like the shirt, or he may like it or we may not like it as much as we thought we would. But does that mean that we don’t use our mental model and simulation for making a decision? No. We all use models all the time, Knowingly or unknowingly.

The real question to be asked is whether we are aware of the fact that we do so. Do we have easy access to the assumptions which go in creation of our mental models? Do we communicate them well to others while making a decision or a policy? What tools can help us elicit our assumptions and create a schema of our mental model so that we can share it with others and so do they?

I believe the answer to the above questions is the use of system dynamics. It is a tool which can help us improve our understanding of our mental models and open it up for others to review and improve. This in turn improves our understanding of the system we are dealing with for effective decision making and policy planning.

Does this mean that we shall find solution to all the problems? No. Does this mean that our model shall be right? No.

But it would allow us to test our assumptions and improve our decision making process. Simulations also give us the power to understand the unintended consequences of our actions. At the end it can improve our confidence levels on the model which we shall use to make decisions.

Further reading on Models and System Dynamics:

http://web.mit.edu/jsterman/www/All_Models_Are_Wrong_(SDR).pdf

http://www.systemdynamics.org/

http://mitsloan.mit.edu/group/system-dynamics/

http://xmile.systemdynamics.org/

Advertisements

4 thoughts on “Why Models are Wrong and still Useful!

  1. Aditi January 10, 2016 / 3:55 pm

    At the outset, many congratulations to you on your presentation and thanks for sharing your perspectives,

    The analogy which you have established successfully communicates your message. However, i have couple of comments,

    1. The system which we are taking here (in your case, may be a city or a state at large) is where models are used to deal with complex variables and multiple variables. As you rightly said, the mental sketch we make while buying say a shirt for someone,is the key factor. Now a lot depends on how old or recent is the sketch. How updated is your mental image and how efficient your mind is in imagining the person, for whom you are buying the shirt.

    Meaning, the assumptions and the inputs. Also the updated inputs is extremely essential.

    2. The problem with models is they run on the basis of certain assumptions and inputs. Now if these are from wrong sources, unauthenticated and incorrect or even not updated for that matter, naturally, the results would not be aligned with the mental sketch/simulations. Like the case of climate models. If the assumptions entered are not appropriate/updated/from authentic sources, the results are often over exaggerated & misleading.

    I know my father requires a 40 size shirt, but off late he has reduced weight and also, he no more wears a formal shirt is a key information that can guide my decision making of buying him a shirt.

    Prima facie, the problem or i would say challenge with models is the database and its efficiency to handle multiple variables to be able to generate a cohesive result.

    So, your explanation is right, but then, what needs to be highlighted in the explanation is while modelling any particular scenario, what needs initial attention rather a lot of attention is the area of assumptions and inputs.

    Closer the assumptions to reality, precise would be the outputs. And that also adds your statement- Do we have easy access to the assumptions which go in creation of our mental models?

    Hope this helps…!! This reminds me of our never ending field trip conversations,..!!

    Good going.

    best
    aditi

    Like

    • Mihir January 10, 2016 / 4:42 pm

      Hi Aditi,

      Thanks for responding. I have picked 6 key things which you have mentioned and I am elaborating on them further:

      1. Now a lot depends on how old or recent is the sketch: This is the problem of updated knowledge or information. It can be dealt with by research, talking to people, reading and all of that. All active models of the world are being constantly updated. But they are still “wrong”.

      2. how efficient your mind is in imagining the person: This relates to the platform and method which one uses for modeling. Also, when you use the word efficient it seems to point more towards the hardware side of us or computers :-). It still does not make them “correct”.

      3. Like the case of climate models. If the assumptions entered are not appropriate/updated/from authentic sources, the results are often over exaggerated & misleading: You have overemphasized on the fact that perfect knowledge exists and rests at some “authentic” sources. That means someone somewhere knows how reality works and has a repository from where we can take the knowledge and then our model will neither over exaggerate or mislead. But who has and where does this perfect knowledge exist? Can it? If not, then how do we model assumptions for which we don’t have empirical basis? A key question in system dynamics fraternity.

      4. challenge with models is the database and its efficiency to handle multiple variables to be able to generate a cohesive result: Yes, again the method and platform one uses for modeling will determine the capacity of the model to generate cohesive result. But cohesive results does not imply accuracy. On the contrary the more cohesive your model and results, the higher their uncertainty. Specially for complex systems.

      5. what needs initial attention rather a lot of attention is the area of assumptions and inputs: Spot on. That’s what determines the nature of the game. Do models highlight them enough? Climate models, economic models, epidemic etc.?

      6. Closer the assumptions to reality, precise would be the outputs: I would stretch our imagination even further. “Closer the assumptions to reality, farther the reality goes”.

      I am finding this conversation very interesting and actually it is helping me to think clearly. So thank you for posting your questions/thoughts.

      Look forward to more rounds.

      Mihir.

      Like

  2. Rajesh January 13, 2016 / 11:57 am

    • I agree that the mental models are not fully right but help taking aware decision, since it has some concrete base.
    • The System Dynamics helps in simulating the future considering certain components of the system, which a person thinks is important and with certain assumptions. We know that all the systems are very complex in nature. The components which we think are important and decisive factors may not be the one. In that case the projections may mislead. So we have to be very careful.
    • There is also a component of sensing which I think the tool does not consider. In that case we should rely in our sensing too.
    • Basically when we play with the system dynamics tool, we are experimenting and trying to figure out the best possible response to the scenario.
    • It helps to reduce the risks of failure or we can say that it helps in taking calculated risks with highest possible leverage.

    Regards
    Rajesh

    Like

    • Mihir January 16, 2016 / 1:12 pm

      Points well taken. But can I ask you to elaborate on some of them? It’s worth more explanation.

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s