Author: Eric

Posted on

Convolutional Neural Networks for Behavioral Cloning – Udacity Self Driving Car Engineer Term 1 Project 3

In the last project we showed, by using python and tensorflow, that we can create a ConvNet to classify traffic signs with a relatively low error rate. What if we want to do something a little more complex, though? Say, something like replicate behaviors that a system should take given a set of inputs.

This is just the problem posed to students with the 3rd project in term 1 of the self driving car engineer nano degree series.

Regression vs Classification

In order to replicate a behavior, we are dealing with a regression problem, which is unlike our previous problem of classification. The main difference being that we are no longer concerned with our network outputs corresponding to a probability that something belongs to a class, in such case it is predicting a discrete class label, but rather, they predict a continuous quantity, such as throttle and steering angles.

If you read the previous post, you will remember that we used a softmax cross-entropy function to convert the outputs of our last layer to probabilities of something belonging to each of our classes. Instead, we will use a mean squared error function for our loss layer, which is then again fed into an adam optimizer.

Project Problem Statement

The goals of this project are the following:

  • Use the provided simulator to collect data of good driving behavior. This includes steering angle, camera imagery, throttle position and brake pressure.
  • Build a convolutional neural network in Keras that predicts steering angles from images.
  • Train and validate the model with a training and validation set.
  • Test that the model successfully drives around the track without leaving the road.

Write up and Results

The model constructed is basically the same as the Nvidia model shown below, with the exception of added dropout layers.

The model used includes ReLU activation layers to introduce nonlinearity, and the data is normalized using Keras’ lambda layer. Additionally, cropping was performed to help normalize images.

Attempts were made to use ELU activation functions, but those proved harder to train and performed worse than when using ReLUs.

In order to prevent overfitting dropout layers were incorporated. Additionally, the model was trained and validated on different data sets to prevent overfitting.

The model was trained using an Adam optimizer with a learning rate of 0.0001.

Training data was chosen that I thought would give the best probability of keeping the vehicle on the road. I used a combination of center lane driving, recovering from left and right side departures from the road, driving around a second track, and spending additional time connecting data from very sharp, “extreme” turns.

After collection of training data I wrote a python sript to normalize the distribution of angles that are fed into the training script. I divided the range of [-1.0 rad, 1.0 rad] into 21 bins and thresholded the number allowed in these bins to 1000. After a lot of data recording the training data distribution appeared as follows:

As is the case with many machine learning problems using neural networks, my goal was to use a previously existing model architecture and re-train it for my application.

My first handful of attempts at training a model indicated that overfitting was occurring as the training set showed a low mean squared error, but the validation MSE was quite high. To combat this I used dropout layers as indicated above.

After final construction of the model and testing with my own data set, the CNN still showed it was having a difficult time keeping the car on track. As with most neural network tasks, it seems that the distribution of the data set is more important than the actual format of the network. I therefore created a script to bin the steering angles into 21 bins and threshold the acceptable number of items in each bin to 1000. After quite a bit of data collection and binning I was able to have 21 bins of nearly 1000 items each, which produced a relatively flat distribution. With this new data set of nearly uniform distribution the trained model seemed to perform quite well and kept the car on track.

The mean squared error loss during training can be seen below:

Keras Implementation of the Nvidia CNN Model

def nvidia():

	shape = (160,320,3)
	model = Sequential()
	model.add(Lambda(lambda x: x / 255.0 - 0.5, input_shape = shape))
	model.add(Cropping2D(cropping = ((50,20),(0,0))))
	model.add(Convolution2D(24,5,5, subsample=(2,2), activation = 'relu'))
	model.add(Convolution2D(36,5,5, subsample=(2,2), activation = 'relu'))
	model.add(Convolution2D(48,5,5, subsample=(2,2), activation = 'relu'))
	model.add(Convolution2D(64,3,3, activation = 'relu'))
	model.add(Convolution2D(64,3,3, activation = 'relu'))
	model.add(Dropout(0.5))
	model.add(Flatten())
	model.add(Dense(100))
	model.add(Dropout(0.5))
	model.add(Dense(50))
	model.add(Dense(10))
	model.add(Dense(1))

	return model
Posted on

Using Convolutional Neural Networks to Classify Traffic Signs – Udacity Self Driving Car Engineer Term 1 Project 2

The second project in the computer vision oriented term 1 tasks the student to train a classifier to correctly classify signs using a convolutional neural network constructed in python using TensorFlow. My full github repo for the project is located here.

The primary CNN model used to solve this classification problem is a modified LeNet architecture with the addition of dropout layers between fully connected layers to prevent overfitting.

CNN Architecture

The standard LeNet-5 architecture is shown below, which is retrieved from the original paper.

LeNet-5 Architecture

The final architecture used is summarized below:

LayerDescription
Input32x32x3 RGB image
Convolution 5×51×1 stride, valid padding, outputs 28x28x6
RELU
Max pooling2×2 stride, outputs 14x14x6
Convolution 5×51×1 stride, valid  padding, outputs 10x10x16
RELU
Max pooling2×2 stride, outputs 5x5x16
Fully connected400 inputs, 120 outputs
RELU
DropoutKeep prob = 0.5
Fully connected120 inputs, 84 outputs
RELU
DropoutKeep prob = 0.5
Fully Connected84 inputs, 43 outputs
Softmax

CNN Building Blocks

Neurons

As with standard neural networks, at the core of the convolutional neural network are still neurons, connected by synapses, which compute a dot product of inputs and weights, add to it a bias, pass this to an activation function, and then output this to the next layer.

Image Courtesy: http://cs231n.github.io/

Convolutional Layer

As you may expect, the convolutional layer is the core building block of convolutional neural networks. The convolutional layer consists of a set of learnable filters, or kernels, that are convolved with the input, in this case a 3 channel image.

During the forward pass, each kernel (there may be more than one), is convolved spatially across the input image, thereby creating a 2-dimensional activation map of that kernel. This results in the network learning a kernel (filter) that will activate when it detects a specific type of feature at a certain spatial position in the input image.

Local Connectivity

Due to the high dimensionality of images, if we were to connect every neuron in one volume to everyone neuron in the next, we would have an almost crazy number of parameters, which would result in a very high computational expense. CNNs therefore depend on the concept of local connectivity, and receptive field. The receptive field, to put it simply, is the size of the kernel used in convolution, which results in only local spatial connections between layers.

Activation Functions

Rectified linear units, or ReLUs, were used as activation functions for the traffic sign classifier CNN. When selecting an activation function, the designer should note that only nonlinear activation functions allow neural networks to compute nontrivial problems using only a small number of nodes. In fact, when a nonlinear activation function is used, then a two-layer neural network can be proven to be a universal function approximator.

Exponential Linear Unit (ELU)

Rectified Linear Unit (ReLU)

Hyperbolic Tangent Function

Logistic Sigmoid

Softmax

The softmax is used in the last fully connected layer to be able to convert outputs from the previous layer into probabilities for each output class. Mathematically, it may be defined as follows:

The ReLU is often preferred to other nonlinear activation functions because it trains the neural network several times faster without a significant penalty to the generalization accuracy.

Alternative nonlinear activation functions that are sometimes used include the hyperbolic tangent function, the exponential linear unit, and the logistic sigmoid function. The ELU is a function that tends to converge cost to zero fast and produce accurate results. The ELU is very similar to the ReLU, except that negative inputs result in a non-zero activation that smoothly becomes equal to -\alpha.

Pooling Layers

Pooling layers act to non-linearly downsample the input image. This is necessary because, at their core, neural networks act to reduce the dimensionality of their inputs; for classification afterall, we need to go from an input image of mxn pixels, with a depth of 3, into a certain class, which is a single output. In other words, pooling layers combine the outputs of neuron clusters in the previous layer into a single neuron input in the next layer.

Max pooling is one of the more common types of pooling functions used. In essence, they downsample by extracting the maximum value in a certain filter space. The image below, taken from wikipedia, illustrates how this is performed for a max pooling filter of dimensionality 2×2 and stride of 2.

Fully Connected Layers

Like the name suggests, fully connected layers connect every neuron in one layer to every neuron in the next layer. Fully connected layers typically appear at the end of a network and serve as the final, high-level reasoning device within CNNs.

The output from the convolutional, pooling, and other layers in a CNN represent high-level features of an input image. It is the job of the fully connected layer to use these features to classify the input image into the appropriate classes based on the training data.

Loss Layers

I mentioned the softmax function above, which is one example of a loss function used in loss layers.

In the traffic sign classifier problem I utilized a softmax cross entropy loss function as the loss operation to be minimized.

Utilizing Dropout to Prevent Overfitting

Dropout is simply a regularization technique that aims to prevent overfitting by randomly, or otherwise, dropping out units in a neural network.

Writeup and Results

The student is provided pickled data that contains a dictionary with 4 key/value pairs:

  • 'features' is a 4D array containing raw pixel data of the traffic sign images, (num examples, width, height, channels).
  • 'labels' is a 1D array containing the label/class id of the traffic sign. The file signnames.csv contains id -> name mappings for each id.
  • 'sizes' is a list containing tuples, (width, height) representing the original width and height the image.
  • 'coords' is a list containing tuples, (x1, y1, x2, y2) representing coordinates of a bounding box around the sign in the image. THESE COORDINATES ASSUME THE ORIGINAL IMAGE. THE PICKLED DATA CONTAINS RESIZED VERSIONS (32 by 32) OF THESE IMAGES

To train the model, I used a learning rate of 0.001, 100 epochs, although I could have used far less, and a batch size of 128.  I chose to use the Adam optimizer because from the papers I’ve read adaptive optimizers appeared to have the best performance and were the least computationally expensive.

My final model results were:

  • training set accuracy of 0.998
  • validation set accuracy of 0.954
  • test set accuracy of 0.939

f an iterative approach was chosen:

  • What was the first architecture that was tried and why was it chosen?
    • I first chose the basic LeNet architecture.
  • What were some problems with the initial architecture?
    • The initial training sets were achieving training accuracies of nearly 1.000 while the validation accuracy was only around 0.870.  This indicated that the model wasn’t able to generalize what it learned in the training set to the validation set well.
  • How was the architecture adjusted and why was it adjusted? Typical adjustments could include choosing a different model architecture, adding or taking away layers (pooling, dropout, convolution, etc), using an activation function or changing the activation function. One common justification for adjusting an architecture would be due to over fitting or under fitting. A high accuracy on the training set but low accuracy on the validation set indicates over fitting; a low accuracy on both sets indicates under fitting.
    • I moved on to try adding layers of convolutions as well as in the fully connected classifier layer.  I then moved on to try some inception modules. From there improved the training data set and added dropout because the training set was achieving very good accuracies whereas the validation set was still achieving relatively poor performance.
  • Which parameters were tuned? How were they adjusted and why?
    • I played with the learning rate a bit but decided to leave it at 0.001.  I increased the number of epochs based on where I saw the optimizer begin to stall.
  • What are some of the important design choices and why were they chosen? For example, why might a convolution layer work well with this problem? How might a dropout layer help with creating a successful model?
    • I think the dropout layers helped because it allowed the model to have backup methods of classification which further allowed the model to generalize to the validation set.

Here is an exploratory visualization of the data set. It is a bar chart showing how the training samples are distributed.

Number of training examples = 34799
Number of testing examples = 12630
Number of validation examples = 4410
Image data shape = (32, 32, 3)
From the histogram we can see there are 43 classses

One thing you’ll notice about the training data set is that some classes are over-represented as compared to others. With this observation I opted to generate additional data to ensure that the distribution was more even. To add more data to the the data set, I translated images, rotated them and applied an affine transformation to shear them.

My final training set had 146574 images (after modified images being added). My validation set and test set had 4410 and 12630 images.  These last two numbers were unaltered because they were given to us as separate pickles. Interestingly when I used train_test_split to generate validation images instead of using the ones given to us I achieved higher validation rates, which makes me question the validation image set that was given to us.

Here are some examples of modified images that were added to the data set:

And here is the histogram of how many images there are after processing for each class:

Here is a random sampling of data from the set of non-processed and processed imagery fed to the network for training:

schikit learn was used to shuffle the data, and tensorflow was used as the primary machine learning library.

Model Architecture

from sklearn.utils import shuffle
import tensorflow as tf
from tensorflow.contrib.layers import flatten

def LeNet(x):    
    # Hyperparameters
    mu = 0
    sigma = 0.1
    
    # SOLUTION: Layer 1: Convolutional. Input = 32x32x1. Output = 28x28x6.
    conv1_W = tf.Variable(tf.truncated_normal(shape=(5, 5, 3, 6), mean = mu, stddev = sigma))
    conv1_b = tf.Variable(tf.zeros(6))
    conv1   = tf.nn.conv2d(x, conv1_W, strides=[1, 1, 1, 1], padding='VALID') + conv1_b

    # SOLUTION: Activation.
    conv1 = tf.nn.relu(conv1)

    # SOLUTION: Pooling. Input = 28x28x6. Output = 14x14x6.
    conv1 = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')

    # SOLUTION: Layer 2: Convolutional. Output = 10x10x16.
    conv2_W = tf.Variable(tf.truncated_normal(shape=(5, 5, 6, 16), mean = mu, stddev = sigma))
    conv2_b = tf.Variable(tf.zeros(16))
    conv2   = tf.nn.conv2d(conv1, conv2_W, strides=[1, 1, 1, 1], padding='VALID') + conv2_b
    
    # SOLUTION: Activation.
    conv2 = tf.nn.relu(conv2)

    # SOLUTION: Pooling. Input = 10x10x16. Output = 5x5x16.
    conv2 = tf.nn.max_pool(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')

    # SOLUTION: Flatten. Input = 5x5x16. Output = 400.
    fc0   = flatten(conv2)
    
    # SOLUTION: Layer 3: Fully Connected. Input = 400. Output = 120.
    fc1_W = tf.Variable(tf.truncated_normal(shape=(400, 120), mean = mu, stddev = sigma))
    fc1_b = tf.Variable(tf.zeros(120))
    fc1   = tf.matmul(fc0, fc1_W) + fc1_b
    
    # SOLUTION: Activation and dropout.
    fc1    = tf.nn.relu(fc1)
    fc1  = tf.nn.dropout(fc1, keep_prob)

    # SOLUTION: Layer 4: Fully Connected. Input = 120. Output = 84.
    fc2_W  = tf.Variable(tf.truncated_normal(shape=(120, 84), mean = mu, stddev = sigma))
    fc2_b  = tf.Variable(tf.zeros(84))
    fc2    = tf.matmul(fc1, fc2_W) + fc2_b
    
    # SOLUTION: Activation and dropout
    fc2    = tf.nn.relu(fc2)
    fc2    = tf.nn.dropout(fc2, keep_prob)

    # SOLUTION: Layer 5: Fully Connected. Input = 84. Output = 43.
    fc3_W  = tf.Variable(tf.truncated_normal(shape=(84, 43), mean = mu, stddev = sigma))
    fc3_b  = tf.Variable(tf.zeros(43))
    logits = tf.matmul(fc2, fc3_W) + fc3_b
    
    return logits



x = tf.placeholder(tf.float32, (None, 32, 32, 3))
y = tf.placeholder(tf.int32, (None))
keep_prob = tf.placeholder(tf.float32) 
one_hot_y = tf.one_hot(y, 43)

rate = 0.001
EPOCHS = 100
BATCH_SIZE = 128

logits = LeNet(x)
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits, one_hot_y)
loss_operation = tf.reduce_mean(cross_entropy)
optimizer = tf.train.AdamOptimizer(learning_rate = rate)
training_operation = optimizer.minimize(loss_operation)

correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(one_hot_y, 1))
accuracy_operation = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
saver = tf.train.Saver()

def evaluate(X_data, y_data):
    num_examples = len(X_data)
    total_accuracy = 0
    sess = tf.get_default_session()
    for offset in range(0, num_examples, BATCH_SIZE):
        batch_x, batch_y = X_data[offset:offset+BATCH_SIZE], y_data[offset:offset+BATCH_SIZE]
        accuracy = sess.run(accuracy_operation, feed_dict={x: batch_x, y: batch_y, keep_prob: 1.0})
        total_accuracy += (accuracy * len(batch_x))
    return total_accuracy / num_examples

Model Training and Testing

# train
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    num_examples = len(X_train)

    print("Training...")
    print()
    for i in range(EPOCHS):
        X_train, y_train = shuffle(X_train, y_train)
        for offset in range(0, num_examples, BATCH_SIZE):
            end = offset + BATCH_SIZE
            batch_x, batch_y = X_train[offset:end], y_train[offset:end]
            sess.run(training_operation, feed_dict={x: batch_x, y: batch_y, keep_prob: 0.5})

            
        training_accuracy = evaluate(X_train,y_train)
        validation_accuracy = evaluate(X_validation, y_validation)
        print("EPOCH {} ...".format(i+1))
        print("Training Accuracy = {:.3f}".format(training_accuracy))
        print("Validation Accuracy = {:.3f}".format(validation_accuracy))
        print()

    saver.save(sess, 'model')
    print("Model saved")

A validation set can be used to assess how well the model is performing, which I mentioned I used scikit learn to split up for me previously. The first ten epochs of training resulted in the following training and validation accuracies:

EPOCH 1 ...
Training Accuracy = 0.678
Validation Accuracy = 0.617

EPOCH 2 ...
Training Accuracy = 0.849
Validation Accuracy = 0.791

EPOCH 3 ...
Training Accuracy = 0.913
Validation Accuracy = 0.867

EPOCH 4 ...
Training Accuracy = 0.940
Validation Accuracy = 0.895

EPOCH 5 ...
Training Accuracy = 0.957
Validation Accuracy = 0.905

EPOCH 6 ...
Training Accuracy = 0.966
Validation Accuracy = 0.925

EPOCH 7 ...
Training Accuracy = 0.974
Validation Accuracy = 0.927

EPOCH 8 ...
Training Accuracy = 0.977
Validation Accuracy = 0.935

EPOCH 9 ...
Training Accuracy = 0.981
Validation Accuracy = 0.935

EPOCH 10 ...
Training Accuracy = 0.985
Validation Accuracy = 0.940

As we can see, the training and validation accuracies are still increasing, meaning beneficial learning is occurring. If we notice, however, that the validation accuracy begins to drop, it is evident that overfitting is occurring and the model is not able to generalize beyond its training data set.

Posted on

Finding Lane Lines on the Road – Udacity Self-Driving Car Engineer Nanodegree Term 1 Project 1

The Udacity SDC Engineer Nanodegree Term 1 begins with using basic computer vision algorithms to detect lane lines from a video and display them on screen.

Problem Statement

In this project, you will use the tools you learned about in the lesson to identify lane lines on the road. Once you have a result that looks roughly like “raw-lines-example.mp4”, you’ll need to get creative and try to average and/or extrapolate the line segments you’ve detected to map out the full extent of the lane lines. You can see an example of the result you’re going for in the video “P1_example.mp4”. Ultimately, you would like to draw just one line for the left side of the lane, and one for the right.

The tools you have are color selection, region of interest selection, grayscaling, Gaussian smoothing, Canny Edge Detection and Hough Tranform line detection. You are also free to explore and try other techniques that were not presented in the lesson. Your goal is piece together a pipeline to detect the line segments in the image, then average/extrapolate them and draw them onto the image for display (as below).

Solution and Write up

I will copy and paste sections from the iPython notebook that I made for this project. The Full project can be found on my github.

1. Describe your pipeline.

My lane line detection pipeline consists of the following steps:

  1. Convert each video frame to grayscale.
  2. Threshold the grayscale image.
  3. Apply region of interest to where lane lines are expected to be.
  4. Perform Canny edge detected (dilated here for easier viewing).

5. Perform the Hough transform (here with lines drawn).

6. Extend hough transform lines and bin them by positive and negative slopes:
Extending is done by taking the two tuples (x1,y1), (x2,y2) output from houghLinesP(), determining the slope for each line, and then determining a leftmost and rightmost point across the image to draw a line between.

7. Determine if both lane lines are detected, if one is not, draw one that is offset from the other lane line by a nominal amount so it is a function of the detected line.
8. If both lane lines are detected, perform least squares fit on positive and negative sloped lines:

9. Draw these least squares fit lines on an empty array and perform region of interest trimming to get rid of line segments extending beyond the horizon.
10. Overlay these lane lines on the original image and display them:

2. Identify potential shortcomings with your current pipeline

Potential shortcomings arise from situations where lane lines do not have high contrast with their background, when lines are occluded, or they are shadowed.  For these situations there is no recovery behavior and vision can only be of so much assistance.

Another shortcoming is that if lane lines become too misaligned from the slopes expected they will not be registered as lane lines due to the binning process I, have which sorts things by their determined slope.

3. Suggest possible improvements to your pipeline

One large improvement would be to low pass or kalman filter the positions and slopes of the lane lines so that they are not memoryless like they currently are.  This would dampen (add inertia) to the system and cause the lines to jitter much less.

Another potential improvement could be to filter the color image according to an RGB mask, as opposed to my simple grayscale threshold.

A further improvement would be to monitor the residual error terms or the covariance of the least squares line fitting to determine how much certainty there is in the fit.  This could be used to monitor the state of the detection algorithm, and if the covariance becomes large some kind of recovery behavior within the code could be triggered.

Posted on

The Innovation Equation – Modelling Corporate Structure and Innovation

The March – April 2019 edition of the Harvard Business Review contains a short article entitled “The Innovation Equation” By Safi Bahcall, in which he tries to explain how it isn’t only culture that can change a company’s willingness to innovate, but also how the incentives within the company change as it grows.

Bahcall notes that many companies like GE, Nokia, RIM (Blackberry) went from being known as being extremely innovative and fun to, well, not. He sought to create a model that would explain how companies like these could so quickly shift from nurturing “crazy” projects – the “loonshots” that transform industries – to rejecting important innovations. In his search to explain this phenomena, he began to realize that it seemed to mimic the kinds of phase transitions that are seen in chemistry and materials science, wherein a sudden change occurs in the collective behavior of the many interacting parts of a system. In water this occurs at it’s freezing point.

In the Innovation Equation Bahcall shows that there is a certain size “at which human groups shift from embracing radical ideas to squashing them,” he calls this the number M. This number M, the number of people at which a corporate innovation phase transition occurs,is not fixed, but is a function of two competing forces: stake in outcome, and the perks of rank.

The numerator of this equation, the stake in outcome, encompasses Equity Fraction, E, Fitness Ratio, F, and Management Span, S. The perks of rank are largely modeled by Salary Growth, G. When put all together, as E, F, and S are increased so too does the number of people at which a transition occurs. When G is increased the number of people required for a transition decreases, as shown below.

\begin{aligned}  1) \displaystyle \qquad M = \frac{ \left (E \cdot S^2 \cdot F \right )}{G}  \end{aligned}

The author uses the following example to illustrate what is being modelled: “Imagine that you’re a designer at a medical device company, and your job is to develop a better pacemaker. It’s 4 PM, and you need to decide how you’ll spend the final hour of the workday. Should you experiment a little more with your design, or should you use the time to network, currying favor with your boss or other influential managers? In other words, should you focus on project work or politics? Such daily choices, faced by pacemaker designers and midlevel workers of all kinds, are what really determine the level of innovation at a company – not cultural changes instituted from the top.

Bellow is a graphical illustration of equation 1, where it can be seen that Equity Fraction, Fitness Ratio and Management Span all drain into M, thereby increasing it, while Salary Growth takes away from M, thereby reducing it.

Pretty Pictures!

Equity Fraction, E

Equity fraction represents the extent to which incentives reflect the outcome of projects as opposed to rank within the organization. Equity fraction directly ties your pay to the quality of your work. Equity comes in two forms however, hard and soft. Hard equity is made up by stock options, grants, commissions, bonuses, etc. Soft equity on the other hand is made up by non-financial benefits, such as peer recognition. Relating back to the pacemaker example, regardless of the type of equity, if E is higher it is more likely that designer will spend the extra hour on project work and not politics.

Fitness Ratio, F

Fitness ratio describes the fractional relationship between project-skill fit (PSF) and return on politics (ROP): F = PSF / ROP.
Project skill fit measures the rewards from investing time in your project, whereas return on politics measures the reward from politicking.

If, for example, there is an excellent fit between one’s skills and their project, they would have a very high PSF and they are more likely to spend more time working on it. Ultimately, if this is the case, there would be no need for schmoozing with others, your skilled work would speak for itself. If, however, you are not well suited to the project that has been assigned to you, you would have a low PSF, and the incremental amount of time spent with your project wouldn’t matter much; it would make more sense for you to invest your last hour of the day in politics as it might be the best, or only, way for you to win a promotion. It should be noted that an overmatch of skills an also result in a low PSF. Imagine a very skilled person who is an expert in their field constantly given junior level tasks. They would likely become bored, not interested in doing their work, and would then become more interested in politicking.

The denominator above, the return on politics, is a difficult to measure parameter that all individuals feel. It’s the extent to which lobbying, networking and self-promotion affect promotion decisions. Bahcall uses the following example to illustrate how the ROP may be limited within a company: “Consider two global manufacturers, company A and company B. Each has a California office with three vice presidents and 30 product designers. In both firms, a spot opens for a fourth VP; one of the 30 designers will be selected. Company A is like most firms: the local office will decide who gets promoted. Through the decision making process – which will take nearly a year – those 30 designers will compete to curry favor with the VPs. The return on politics is high. At company B, however, an independent evaluator who has no ties to anyone in the California office will conduct an assessment and present the findings to an independent group of executives who will make the decision. Since there’s little benefit to lobbying, designers at company B will be likely to focus on their projects and on collaborating well. The return on politics is much lower.“

Management Span, S

Management span, also known as span of control, refers to the average number of direct reports that executives in a company have. If a company has a narrow span, and therefore managers have few direct reports, there are many layers to the company’s organizational structure, and promotions are on everyone’s minds. As a result, researchers are going to be more tempted to worry about titles and status than on actual problem solving. If, however, a company has a relatively large span, each manager has many direct reports, which results in few layers to an organizational structure, promotions won’t be on people’s minds as they rarely occur. This results in people focusing more on their work and less on politicking. Bahcall notes that narrow spans are generally better if you want low error rates and high operational excellence, whereas wider spans and looser controls are better for experimenting and developing new technologies.

Salary Growth, G

Salary growth characterizes the average step-up in base salary, and other perks, that employees receive as they ascend the corporate ladder. For example, if every step up the ladder also came with a 200% increase in salary, you’d try your hardest to make sure that every influential person knew exactly who you were. On the other hand, if each step was accompanied by only a 2% increase in salary, people wouldn’t really care too much.

Low salary step-up rates encourage people to use the last hour of the day on work, not on politicking. One recent academic study even went as far to conclude that “increased [wage] dispersion is associated with lower productivity, less cooperation, and increased turnover.”

What Should You Do?

In order to adjust the control parameters to increase M and enhance innovation Bahcall suggests trying a few different things:

Celebrate results, not rank. To increase the equity fraction and lower the salary growth rate, management must structure rewards to be based more on results than on level in the hierarchy.

Use soft equity. As I talked about in my post about The Five Love Languages, people are motivated by different things, and feel different emotional responses than others to the same experience. Some people may be more motivated by tangible financial rewards. Others may be more driven by peer recognition, or a sense of accomplishment and personal growth. Companies should try their hardest to identify and use all motivational means at their disposal.

Take politics out of the equation. Employees need to see that lobbying for pay and promotions will not help them. “When promotions are considered at McKinsey, for example, a partner from a different office and preferably a different functional practice interviews candidates colleagues and clients and then reports back to a group of partners who make a decision.”

Invest in training.

Perfect employee placement. Designate a person or team to regularly monitor the organization for good skill fit.

Fine tune your management span.

Appoint a chief incentive officer. “Organizations need top-level executives who are well trained in the subtleties of aligning incentives and solely focused on achieving a state-of-the-art compensation system. A good incentives officer can identify wasteful bonuses, reduce the risk of perverse incentives, and tap into the power of nonfinancial rewards. The goal of achieving the most motivated employees for a given compensation budget is as important and strategic to companies as is the goal of achieving the best sales for a given marketing budget (the province of a chief revenue officer) or the best T systems for a given technology budget (a chief information officer’s terrain).”

“Culture still matters, of course, but it’s time to pay a little more attention to structure.”

Posted on

The Five Love Languages

The Five Love Languages by Gary Chapman

I originally put off reading this book for as long as I could, mostly because I kept seeing all the quizzes floating around Facebook offering to “Find your love language!” and the like. Ultimately though, I’m glad I read this book. I think it has more words than it needs, and can pretty accurately be summarized in one sentence. People’s emotional response to actions are often different from yours.

This is obviously a bit of a generalization from the topic of the book, which aims to be more of a relationship self-help book, but I believe it to be quite sound. Chapman aims to show the reader that your expression of love will not be interpreted as “showing love” unless you and your loved one are speaking the same “love language.”

Chapman proposes the following love languages:

  • Words of Affirmation
  • Quality Time
  • Receiving Gifts
  • Acts of Service
  • Physical Touch

Quotes and Excerpts

“The important thing is to speak the love language of your souse. Seldom do a husband and wife have the same primary emotional love language. We tend to speak our primary love language, and we become confused when our spouse does not understand what we are communicating. We are expressing our love, but the message does not come through because we are speaking what, to them, is a foreign language… Love need not evaporate after the wedding, but in order to keep it alive, most of us will have to put forth the effort to learn a secondary love language. We cannot rely on our native tongue if our spouse does not understand it. If we want them to feel the love we are trying to communicate, we must express it in their primary love language.”

This is one that I think everyone needs to hear, because we are often marketed misconceptions and lies by the “love industry”:
“We have been led to believe that if we are really in love, it will last forever. We will always have the wonderful feelings that we have at this moment. Nothing could ever come between us. Nothing will ever overcome our love for each other… We observe that some married coupes seem to have lost that feeling, but it will never happen to us. ‘Maybe they didn’t have the real thing,’ we reason… Unfortunately, the eternality of the ‘in-love’ experience is fiction, not fact… After studying scores of couples, she (Dr. Dorothy Tennov) concluded that the average life span of a romantic obsession is two years. If it is a secretive love affair, it may last a little longer. Eventually, however, we call descend from the clouds and plant our feet on earth again… What happened to the ‘in-love’ experience? Alas, it was but an illusion by which we were tricked into signing our names on the dotted line, for better or for worse. No wonder so many have come to curse marriage and the partner whom they once loved. After all, if we were deceived, we have a right to be angry. Did we really have the ‘real’ thing? I think so. The problem was faulty information. The bad information was the idea that the ‘in-love’ obsession would last forever. We should have known better. A casual observation would have taught us that if people remained obsessed, we would all be in serious trouble.”

“Research seems to indicate that there is a third and better alternative (to resigning yourself to a life of misery with their spouse, or jump ship and try again): We can recognize the in-love experience for what it was – a temporary emotional high – and now pursue ‘real love’ with our spouse.”

Words of Affirmation

Verbal compliments, or words of appreciation, are powerful communicators of love. They are best expressed in simple, straightforward statements of affirmation such as:
‘you look sharp in that suit’
‘Do you ever look incredible in that dress! Wow!’
‘I really like how you’re always on time to pic me up at work.’
‘Thanks for getting the babysitter lined up tonight. I want you to know I don’t take that for granted.’
‘You can always make me laugh.’

“Verbal compliments are far greater motivators than nagging words.” And this makes sense, right? Just think about you inner B.F. Skinner and operant conditioning. I’m going to be much more motivated to do something if I receive praise for it than if I’m nagged for not doing it.

“The object of love is not getting something you want, but doing something for the well-being of the one you love. It is a fact, however, that when we receive affirming words we are for more likely to be motivated to reciprocate and do something our spouse desires.”

“Giving verbal compliments is only one way to express words of affirmation to your spouse. Another dialect is encouraging words… All of us have areas in which we feel insecure. We lack courage, and that lack of courage often hinders us from accomplishing the positive things that we would like to do.”

Dialects of Words of Affirmation

  • verbal compliments
  • encouraging words
  • kind words
    • “Love doesn’t keep a score of wrongs. Love doesn’t bring up past failures… If I choose to forgive, intimacy can be restored. Forgiveness is the way of love.”
    • “I am amazed by how many individuals mess up every new day with yesterday. They insist on bringing into today the failures of yesterday, and in so doing, they pollute a potentially wonderful present.”
    • “The best thing we can do with the failures of the past is to let them be history. Yes, it happened. Certainly it hurt. And it may still hurt, but he has acknowledged his failure and asked your forgiveness. We cannot erase the past, but we can accept it as history.”
  • humble words
    • “Love makes requests, not demands. When I demand things from my spouse, I become a parent and she the child.”
    • “If we wish to love each other, we need to know what the other person wants. The way we express those desires, however, is all-important. If they come across as demands, we have erased the possibility of intimacy and will drive our spouse away.”
    • “When you make a request of your spouse, you are affirming his or her worth and abilities. You are in essence indicating that she has something or can do something that is meaningful and worthwhile to you. When, however, you make demands, you have become not a lover, but a tyrant.”
    • “We cannot get emotional love by way of demand. My spouse may in fact comply with my demands, but it is not an expression of love. It is an act of fear or guilt or some other emotion, but not love.”

“Psychologist William James said that possibly the deepest human need is the need to feel appreciated. Words of affirmation will meet that need in many individuals. If you are not a man or woman of words, if it is not your primary love language but you think it may be the love language of your spouse, let me suggest that you keep a notebook titled Words of Affirmation. When you read an article or book on love, record the words of affirmation you find. When you hear a lecture on love or you overhear a friend saying something positive about another person, write it down. In time, you will collect quite a list of words to use in communicating love to your spouse.”

Quality Time

Chapman defines quality time as “Giving someone your undivided attention… What I mean is sitting on the couch with the TV off, looking at each other and talking, devices put away, giving each other your undivided attention. It means taking a walk, just the two of you, or going out to eat and looking at each other and talking.”

“When I sit with my wife and give her twenty minutes of my undivided attention, and she does the same for me, we are giving each other twenty minutes of our life. We will never have those twenty minutes again; we are giving our lives to each other. It is a powerful emotional communicator of love.”

“I explained the concept of giving someone your undivided attention, not talking to her while you watch sports or read texts but looking into her eyes, giving her your full attention, doing something with her that she enjoys doing and doing it wholeheartedly.”

“Her love language is quality time. You have given her so little time that her love tank is empty. She doesn’t feel secure in your love. Therefore she has lashed out at what was taking your time her mind – your job. She doesn’t really hate your job. She hates the fact that she feels so little love coming from you.”

“A key ingredient in giving your spouse quality time is giving them focused attention, especially in this era of many distractions. When a father is sitting on the floor, rolling a ball to his two-year old, his attention is not focused on the ball, but on his child. For that brief moment, however long it lasts, they are together. If, however, the father is talking on the phone while he rolls the ball, his attention is diluted. Some husbands and wives think they are spending time together when, in reality, they are only living in close proximity. They are in the same house at the same time, but they are not together.”

“Quality time does not mean that we have to spend our together moments gazing into each other’s eyes. It means that we are doing something together and that we are giving our full attention to the other person. The activity in which we are both engaged is incidental. The important thing emotionally is that we are spending focused time with each other. The activity is a vehicle that creates the sense of togetherness.”

Dialects of Quality Time

  • Quality Conversation
    • Sympathetic dialogue where two individuals are sharing their experiences, thoughts, feelings and desires in a friendly, uninterrupted context.
    • Many people fall into the trap of suggesting solutions when someone talks about the misfortunes or frustrations of their day. In fact, many people simply want sympathy, to know that someone understands and validates their feelings. We must be willing to give advice, but only when it is requested and never in a condescending manner.
      • I will keep coming back to this notion of emotional validation in further posts, but know that it is central in the framework of Dialectical Behavior Therapy.
    • TIPS:
      • maintain eye contact when speaking.
      • don’t listen to your spouse and do something else at the same time.
      • listen for feelings. Ask your self, “what emotion is my spouse experiencing?”
      • Observe body language.
      • refuse to interrupt.
    • Quality Activities

Learning to Talk

“Quality conversation requires not only sympathetic listening, but also self-revelation*. When a wife says, ‘I wish my husband would talk. I never know what he’s thinking or feeling,’ she is pleading for intimacy… Self-revelation does not come easy for some of us. We may have grown up in homes where the expression of thoughts and feelings was not encouraged but squelched. To request a toy was to receive a lecture on the sad state of family finances. The child went away feeling guilty for having their desire, and he quickly learned not to express his desires. When he expressed anger, the parents responded with harsh and condemning words. Thus, the child learned that expressing angry feelings is not appropriate. If the child was made to feel guilty for expressing disappointment at not being able to go to the store with his father, he learned to hold his disappointment inside. By the time we reach adulthood, many of us have learned to deny our feelings.**”

* Self-revelation is just a synonym for “Self-Awareness”, or more accurately “Emotional Self-Awareness” within the context of the Emotional Intelligence frameworks presented in other posts.
** Within the context of Dialectical Behavior Therapy this is known as emotional invalidation. We are told we should not feel what we are feeling. It results in our emotional responses just becoming stronger and disallowing the development of the appropriate tools to deal with them.

“Perhaps he has reason to feel angry or disappointed, but he has lived so long in the world world of thought that he does not acknowledge his feelings.”

“Remember, emotions themselves are neither good nor bad. They are simply our psychological responses to the events of life*.

* This is a core concept of mindfulness

“Not all of us are in touch with our emotions, but when it comes to talking, all of us are affected by our personality. I have observed two basic types: Dead Seas and Babbling Brooks. The former so called because it receives but it does not give. This personality type receives many experiences, emotions and thoughts throughout the day. They have a large reservoir where they store that information, and they are perfectly happy not to talk. The latter so called because whatever enters the eye gate or the ear gate comes out the mouth gate, and there are seldom sixty second between the two… Many times a Dead Sea marries a Babbling Brook. That happens because when they are dating, it is a very attractive match. If you are a Dead Sea and date a Babbling Brook, you will never have to think ‘how will I start the conversation tonight?’ ‘how will I keep the conversation flowing?’ In fact, you don’t have to think at all. All you have to do is nod your head and say ‘uh-huh’, and she will fill up the whole evening and you will go home saying ‘what a wonderful person.’ On the other hand, if you’re a Babbling Brook and you date a Dead Sea you’ll have an equally wonderful evening because Dead Seas are the world’s best listeners. You will babble for three hours and he will listen intently to you.”

Receiving Gifts

“A gift is something you can hold in your and and say, ‘Look, he was thinking of me,’ or ‘she remembered me.’ You must be thinking of someone to give him a gift. The gift is a symbol of what you thought. It doesn’t matter whether it costs money. What is important is that you thought of him.”

“Don’t wait for a special occasion. If receiving gifts is his/her primary love language, almost anything you give will be received as an expression of love.”

“Physical presence in the time of crisis is the most powerful gift you can give if your spouse’s primary love language is receiving gifts.”

Acts of Service

“By acts of service, I mean doing things you know your spouse would like you to do. You seek to please her by serving her, to express your love for her by doing things for her… They require thought, planning, time, effort, and energy. If done with a positive spirit, they are indeed expressions of love.”

“They don’t necessarily require a lot of time. One man always dreaded the chore of bringing in the garbage cans from the curb at the end of a long workweek… The empty curb was a clear message to me: ‘I was thinking of you. You were with me, even when you were gone.'”

Physical Touch

“Physical tough is a powerful vehicle for communicating marital love. Holding hands, kissing, embracing, and sexual intercourse are all ways of communicating emotional love to one’s spouse. For some individuals, physical touch is their primary love language. Without it, they feel unloved. With it, their emotional tank is filled, and they feel secure in the love of their spouse.”

“Whatever there is of me resides in my body. To touch my body is to touch me. To withdraw from my body is to distance yourself from me emotionally.”

Determining Your Primary Love Language

  • What does your spouse do or fail to do that hurts you most deeply? The opposite of what hurts you most is probably your love language.
  • What have you most often requested of your spouse? The thing you have most often requested is likely the thing that would make you feel most loved.
  • In what way do you regularly express love to your spouse? Your method of expressing love may be an indication that that would also make you feel loved.

Other Thoughts

“Thousands of husbands and wives have been there – emotionally empty, wanting to do the right thing, not wanting to hurt anyone, but being pushed by their emotional needs to seek love outside the marriage.”

“In time, however, we come down from that natural high back to the real world. If our spouse has learned to speak our primary love language, our need for love will continue to be satisfied. If on the other hand, he or she does not speak our love language, our tank will slowly drain, and we will no longer feel loved… After some years of living with an empty love tank, she will likely ‘fall in love’ with someone else, and the cycle will begin again.”

“Few men who suffer from an empty emotional love tank leave their marriage until they have prospects of meeting that need somewhere else.”

“…Fortunately Brent was the benefactor of what I call the disequilibrium of the ‘in-love’ experience. That is, almost never do two people fall in love on the same day, and almost never do they fall out of love on the same day. You don’t have to be a social scientist to discover that truth. Just listen to country music.”

Posted on

The Importance of Coaches

The importance of coaches, outside of athletic endeavors, was first brought to my attention by someone discussing the article Personal Best by Atul Gawande in the October 2011 issue of The New Yorker’s Annals of Medicine. And it makes sense, right? Athletes, singers, performers of all kinds, have coaches monitoring their every move and providing feedback to further fine tune their behaviors, so why shouldn’t this also be a useful tool in the work place, or even in your interpersonal life?

The trap many people fall into is that they are under the impression that they are able to adequately assess their own actions. They are the ones who know themselves the beset, after all. But, the issue here is that our perceptions are not objective, they are distorted by many things, and in the case of athletics, our sense of proprioception is never going to as good as another set of skilled, watchful eyes.

Take for example my more recent foray into superbike racing. I’ve been riding motorcycles nearly all my life, and finally took the leap to begin road racing. As anyone who has ever watched roadracing knows, the coolest thing you can do is drag knee through a corner. I mean, look at how awesome Marc Marquez looks doing this:

MotoGP World Champion Marc Marquez

Now, when I started my coaching sessions, I knew I was so close to dragging knee… Until I got my coaches feedback and they showed me pictures of myself.

What the hell, I thought. I knew I was so close to dragging body parts on the race track. I could tell I was hanging off the bike so far, I was going so fast. Well, my perception of what was actually happening was far from the objective truth of it. And this is where some people have a very difficult time with things. They come into a situation and fall into one of the biggest traps someone can fall into, regardless of if they’re new or an elite level athlete: they aren’t coachable. They think their perception of reality is the objective truth of it. They think the feedback they are getting is wrong.

Now, for the sake of saving face, I listened to my coaches and got there. I mean, look at how cool I look.

Why This Matters Beyond Athletics

Ultimately, coaching helps to strengthen our self-awareness, emotional or otherwise. We may think we are doing something, whether it’s cornering effectively in motorcycle road racing, or responding adequately to an emotionally charged or difficult conversation in the workplace, but if we are only relying on our own perception of it, we are probably pretty far from the objective truth of the matter.

If we proceed day in and day out by assuming that our interpretations of things are correct,without ever checking that assumption, we will continue to reinforce behaviors based on these perceptions (or inputs). But what if these are wrong? Well, we will begin developing behaviors that are at best much less effective than they could be, and at worst completely inappropriate to the situation.

In Personal Best, Atul Gawande talks about how this has related to his performance as a surgeon. In his earlier years he improved every year, ultimately performing much better than national averages. That is, until one day he plateaued. Gawande recalls being at a medical conference and searching for an impromptu tennis match, having been a very competitive player in high school that took particular prided his serve. During the match, the other player, who played in college, mentioned “you know, you could get more power from your serve.”

He realized that what he thought he was doing, was not in fact what he was actually doing. He gradually realized his legs weren’t really underneath him when we swung his racket up into the air. His right leg dragged a few inches behind his body. With these realizations, and a bit of practice, his serve was drastically improved.

What Is a Coach, Anyway?

The concept of a coach is somewhat difficult to define. They teach, but they are not teachers. They can be bossy, but they’re not really your boss. In fact, they don’t even have to be good at what they’re coaching, they just have to be knowledgeable about it. Coaches simply observe, make a judgement about what they see based on their knowledge of the subject, and then guide.

As Gawande mentions, the coaching model is distinct from the traditional concept of pedagogy; in the latter there’s a presumption that the student will no longer need instruction. You get to a certain point and you’re done. You know everything there is to know. But how does this “being done” notion work in a world where things are always changing, techniques evolve, etc? In my mind, it fails many people.

The model for coaching, as opposed to pedagogy, is different in that it considers the latter naive about our human capacity for self-perfection. It firmly states that no matter how well prepared someone may have been at one point in time, nearly no one can actually maintain their best performance on their own.

Utilizing a Coach

Gawande inquires about the lack of coaching in classical performing arts with Julliard graduate and violin virtuoso Itzhak Perlman only to find out that, in a sense, he actually had a coach. Gawande asked “Why concert violinists didn’t have coaches, the way top athletes did. He (Perlman) said that he didn’t know but that it had always seemed a mistake to him. He had enjoyed the services of a coach all along.” Perlman indicated that he felt incredibly lucky to have met his wife, who was also a concert-level violinist, whom he’d relied on for feedback for the last 40 years. According to Perlman, “The great challenge in performing is listening to yourself. Your physicality, the sensation that you have as you play the violin, interferes with your accuracy of listening. What violinists perceive is often quite different from what audiences perceive. My wife always says that I don’t really know how I play. She is an extra ear… She is very tough, and that’s what I like about it.” Perlman’s wife, all along, had been next to him telling him if a passage was too fast, or too tight, or too mechanical, or any other myriad things that needed fixing.

California researchers in the early 80’s conducted a five year study of teacher-skill development in schools and noticed that when teachers only attended workshops new skill adoption rates were approximately ten percent, but when coaching was introduced, adoption rates passed 90 percent.

Good coaches know how to dissect a performance into its constituent components. They remove little items from the complex behaviors, bring them to the forefront to be thought about, practiced, and then insert them back in to the complex behavior. In sports, coaches focus on mechanics, conditioning, and strategy. In professional settings, and in particular in the classroom as Gawande illustrates, coaches do the same thing, but often focus on behaviors, attention span, and rates of learning.

Gawande decided to try to incorporate this concept of a coach into his operating room. He asked an accomplished surgeon to monitor him during one of his operations, one that he said couldn’t have gone better. Upon meeting with his coach afterward, who observed the entire operation, he was given many, albeit very minor, changes to incorporate. He was told to leave more room to the left which would have allowed the medical student to hold the retractor and free up the surgical assistant’s left hand. He was told to pay more attention to his elbows. He was told to think more about the positioning of the patient in comparison to everyone else, not just him. He was told a few times that he would have been more efficient, and therefore less tired, had he chosen a different instrument a couple of times, or changed his body positioning. Gawande stated “That one twenty minute discussion gave me more to consider and work on than I’d had in the past five years.”

When it comes down to it, all a coach does is monitor your inputs and outputs to a system, and provides a feedback mechanism to optimize the outputs. This often times means changing inputs, whether it’s the mechanics you’re using as an athlete, a particular teaching strategy you use as an instructor, or the manner in which you provide feedback to direct reports as a manager.

An Engineering Systems Example

Think about it like this: if the speedometer on your car is off by 20%, the speed you go when you set your cruise control will be wrong, even if you, and the car, both think it’s correct. Ultimately though, what you and your car think doesn’t matter as much as the objective reality of the situation when you get a speeding ticket. In this scenario the system is your car, including the cruise control, the input is your desired speed, and the output is the actual speed. The feedback by which the car is able to achieve this output speed is from some variety of wheel speed sensor or calibrated tachometer. The police officer is the coach.

In my motorcycle example above, the system is me and my motorcycle, the input is the lean angle and body position I want, and the output is the lean angle and body position that I’m achieving. My feedback mechanism is my internal proprioception. The coach is, well, my coach.

Some Final Thoughts

Coaching outside of athletics has become a bit of a fad over the last handful of years, but beware that bad coaching will make people worse. Coaches should foster an effective solution or change, not merely have you replicate a technique that they like. They should help you find what works for you, given that you and your situation are not just an exact duplicate of another thing like machines are.

Posted on

Emotional Intelligence Has More Facets Than Most People Think

In the February 2017 edition of the Harvard Business Review, Daniel Goleman and Richard Boyatzis wrote a short article entitled Emotional Intelligence has 12 Elements. Which do you need to work on? that I not only thought was interesting, but captured a lot of the misconceptions that I see about emotional intelligence.

They discuss a manager of a small team who is very well liked, kind, respectful, sensitive to the needs of others. She is always engaged and is a source of calm to her colleagues. She’s a good problem solver, and tends to see setbacks as opportunities. Despite her boss complimenting her on her high levels of EI, and her even considering EI as a strength of hers, she feels stuck in her career, unable to demonstrate the kind of performance her company is looking for. She thinks to herself so much for emotional intelligence.

The trap that she’s found herself, and her manager, in is that they are defining emotional intelligence much too narrowly. They’re thinking of emotional intelligence as only sociability, sensitivity and likability. They’re missing things that are important but overlooked by many: the ability to deliver difficult feedback to employees, the courage to ruffle feathers and drive change, and the creativity to think outside the box. These are examples of a case when someone has uneven emotional intelligence skills, or emotional competencies. Goleman states that people often have an imbalance in their EI skills, especially those that are viewed as having high EI, but having a well-balanced array of specific EI capabilities actually prepares a leader for exactly the kinds of tough challenges as those listed above.

There are a number of different models for the different aspects of EI, and even I have a different “5 Aspect” model shown above, but Goleman prefers to use a four domain, 12 aspect model as shown below.

Goleman and Boyatzis Emotional Intelligence Model

G and B, as I shall refer to them, prefer to use a model of emotional intelligence that utilizes the four domains of Self-awareness, self-management, social-awareness, and relationship management. Within these four domains they define twelve EI competencies that are learned and learnable capabilities that allow outstanding performance at work or as a leader. To relate back to their original example, our small-team manager is strong in empathy, positive outlook, and self-control, but she appears to be lacking in conflict management, influence, and inspirational leadership, which are skills that require just as much engagement with emotions as her strengths, and should be worked on just the same as other EI competencies.

If one has strength in conflict management they will be skilled in giving people unpleasant feedback. If one is more inclined to influence the group, they would also want to provide that difficult feedback as a way to lead her direct reports and help them grow.

The next example they give I love, and hope a friend of mine reads this as it’s pretty related to a conversation we were having yesterday about EI (*cough* Annie *cough*). “Say, for example, that Esther has a peer who is overbearing and abrasive (or has any other quality that is against the House Style of the organization). Rather than smoothing over every interaction, with a broader balance of EI skills she could bring up the issue to her colleague directly, drawing on emotional self-control to keep her own reactivity at bay while telling him what, specifically, does not work in his style. Bringing simmering issues to the surface goes to the core of conflict management. Esther could also draw on influence strategy to explain to her colleague that she wants to see him succeed, and that if he monitored how his style impacted those around him he would understand how a change would help everyone.”

G and B conclude by stating the importance of taking a comprehensive “formal 360-degree assessment, which incorporates systematic, anonymous observations of your behavior by people who work with you, have been found to not correlate well with IQ or personality, but they are the best predictors of a leader’s effectiveness, actual business performance engagement, and job (and life) satisfaction.” One thing that is worth noting here, is this is just another example of the importance of coaching in professional and personal life. I will discuss this later, but coaching goes far beyond the athletic world, and should be used as frequently as possible. It is very easy for someone to think they are doing one thing, only to be told by a competent second set of eyes that they are doing far from what they think they are.

Posted on

Emotional Intelligence pt. I – The Emotional Brain

Emotional Intelligence by Daniel Goleman

I can think of no better way to start talking about emotional intelligence than by discussing the book that popularized the topic. Written by author and science journalist Daniel Goleman, Emotional Intelligence analyzes the importance of EI in every aspect of life. Goleman posits that EI is as important, if not more so, than IQ in determining success in academic, professional, social and interpersonal aspects of life. Goleman argues that EI is a skill that can be taught and cultivated, and he proceeds to outline methods for incorporating emotional skills training in school curriculum.

The first part of the book discusses the biological underpinnings of emotions, where they originate in the brain, and why we have them. Ultimately, emotions are our internal messaging system, whose evolution over the last 50,000 “human generations” have been finely tuned to keep us alive in the environment in which we evolved; small tribes. The quick departure from the state in which we evolved due to the rise of civilization has caused an increasing mismatch in the suitability of these emotions in our daily lives. Emotional Intelligence, and the corresponding Emotional Competence, serves as an ever more necessary skill set to help us deal with these mismatches, communicate and relate to others, and succeed in nearly every aspect of daily life.

Quotes and Excerpts

Introduction

“…But IQ washes out when it comes to predicting who, among a talented pool of candidates within an intellectually demanding profession, will become the strongest leader… EI abilities rather than IQ or technical skills emerge as the ‘discriminating’ competency that best predicts who among a group of very smart people will lead most ably.”

“At the very highest levels, competence models for leadership typically consist of anywhere from 80 to 100 percent IE based abilities. As the head of research at a global executive search firm put it. “CEOs are hired for their intelligence and business expertise – and fired for a lack of emotional intelligence.”

“While our emotional intelligence determines our potential for learning the fundamentals of self-mastery an the like, our emotional competence shows how much of that potential we have mastered in ways that translate to on the job capabilities.”

Aristotle’s Challenge

“What factors are at play, for example, when people of high IQ flounder and those of modest IQ do surprisingly well? I would argue that the difference quite often lies in the abilities called here emotional intelligence, which includes self-control, zeal and persistence, and the ability to motivate oneself. And these skills, as we shall see, can be taught to children, giving them a better chance tot use whatever intellectual potential the genetic lottery may have given them.”

“There is growing evidence that fundamental ethical stances in life stem from underlying emotional capacities. For one, impulse is the medium of emotion; the seed of all impulse is a feeling bursting to express itself in action. Those who are at the mercy of impulse – who lack self control – suffer a moral deficiency: The ability to control impulses is the base of will and character. By the same token, the root of altruism lies in empathy, the ability to read emotions in others; lacking a sense of another’s need or despair, there is no caring. And if there are any two moral stances that our times call for, they are precisely these. self-restraint and compassion.”

Note: I will address this in a future post, but according to models within dialectical behavior therapy there is MUCH more to just “impulse control”, as Goleman somewhat naively and succinctly puts it.

“Part two of this book is in seeing how neurological givens play out in the basic flair for living called emotional intelligence: being able, for example, to rein in emotional impulse; to read another’s innermost feelings; to handle relationships smoothly – as Aristotle put it, the rare skill ‘to be angry with the right person, to the right degree, at the right time, for the right purpose, and in the right way… As Aristotle saw, the problem is not with emotionality, but within the appropriateness of emotion and its expression.'”

1 – What Are Emotions For?

“…That power is extraordinary: Only a potent love – the urgency of saving a cherished child – could lead a parent to override the impulse for personal survival. Seen from the intellect, their self-sacrifice was arguably irrational; seen from the heart, it was the only choice to make. Sociobiologists point to the preeminence of heart over head at such crucial moments when they conjecture about why evolution has given emotion such a central role in the human psyche. Our emotions, they say, guide us in facing predicaments and tasks too important to leave to intellect alone – danger, painful loss, persisting toward a goal despite frustrations, bonding with a mate, building a family.”

“But while our emotions have been wise guides in the evolutionary long run, the new realities civilization presents have arisen with such rapidity that the slow march of evolution cannot keep up… In terms of biological design for the basic neural circuitry of emotion, what we are born with is what worked best for the last 50,000 human generations, not the last 500 generations – and certainly not the last five. The slow, deliberate forces of evolution that have shaped our emotions have done their work over the course of a million years; the last 10,000 years – despite having witnessed the rapid rise of human civilization ad the explosion of the human population from five million to five billion – have left little imprint on our biological template for emotional life.”

“All emotions are. in essence, impulses to act, the instant plans for handling life that evolution has instilled in us… That emotions lead to actions is most obvious in watching animals or children; it is only in ‘civilized’ adults we so often find the great anomaly in the animal kingdom, emotions – root impulses to act – divorced from obvious reaction.”

“In our emotional repertoire each emotion plays a unique role…
Anger – blood flows to the hands, making it easier to grasp a weapon or strike at a foe; heart rate increases, and a rush of hormones such as adrenaline generates a pulse of energy strong enough for vigorous action.
Fear – blood goes to the large skeletal muscles making it easier to flee – and making the face blanch as blood is shunted away from it. Circuits in the brain’s emotional centers trigger a flood of hormones that put the body on general alert, making it edgy and ready for action, and attention fixates on the threat at hand.
Happiness – increased activity in a brain center that inhibits negative feelings and fosters an increase in available energy, and a quieting of those that generate worrisome thought.
Love –
Surprise –
Disgust –
Sadness – A main function of sadness is to help adjust to a significant loss. It brings a drop in energy and enthusiasm for life’s activities, particularly diversions and pleasures… This lose of energy may well have kept saddened – and vulnerable – early humans close to home, where they were safer.”

“Those same pressures had made our emotional responses so valuable for survival; as they waned, so did the goodness of fit of parts of our emotional repertoire. While in the ancient past a hair-trigger anger may have offered a crucial edge for survival, the availability of technology that can harm others has made it too often a disastrous reaction.”

“In a very real sense we have two minds, one that thinks and one that feels.”

“From the most primitive root, the brainstem, emerged the emotional centers. Millions of years later in evolution, from these emotional areas evolved the thinking brain or neocortex, the great bulb of convoluted tissues that make up the top layers. The fact that the thinking brain grew from the emotional reveals much about the relationship of thought to feeling; there was an emotional brain long before there was a rational one.”

“With the arrival of the first mammals came new, key layers of the emotional brain. These, surrounding the brainstem, look roughly like a bagel with a bite taken out at the bottom here the brainstem nestles into them. Because this part of the brain rings and borders the brainstem, it was called the ‘limbic’ system, from limbus, the latin word for ring. This new neural territory added emotions proper to the brain’s repertoire. When we are in the grip of craving or fury, head over heels in love or recoiling in dread, it is the limbic system that has us in its grasp… As it evolved, the limbic system refined two powerful tools: learning and memory.”

“The homo sapien neocortex, so much larger than in any other species, has added all that is distinctly human… Our survival edge is due to the neocortex’s talent for strategizing, long-term planning, and other mental wiles. Beyond that, the triumphs of art, of civilization and culture, are all fruits of the neocortex… As we proceed up the phylogenetic scale from reptile to rhesus to human, the sheer mass of the neocortex increases. With that increase comes a geometric rise in the interconnections in brain circuitry. The larger the number of such connections, the greater the range of possible responses. The neocortex allows for the subtlety and complexity of emotional life, such as the ability to have feelings about feelings. There is more neocortex to limbic system in primates than in other species – and vastly more in humans – suggesting why we are able to display a far greater range of reactions to our emotions, and more nuance.”

2 – Anatomy of an Emotional Hijacking

Chapter two starts out with a story about a serial burglar who in a fit of panic clubbed two women over the head until they were unconscious, then completely engulfed by rage and fear murdered them with a kitchen knife. The burglar stated that “I just went bananas. My head just exploded.”

“Such emotional explosions are neural hijackings. At those moments, evidence suggests, a center of the limbic brain proclaims an emergency, recruiting the rest of the brain to its urgent agenda. The hijacking occurs in an instant, triggering this reaction crucial moments before the neocortex, the thinking brain, has had a chance to glimpse fully what is happening, let alone decide if it’s a good idea. The hallmark of such a hijack is that once the moment passes, those so possessed have the sense of not knowing what came over them… Neural takeovers, as we shall see, originate in the amygdala, a center in the limbic brain.”

“The hippocampus and the amygdala were the two key parts of the primitive “nose brain” that, in evolution, gave rise to the cortex and then the neocortex. To this day these limbic structures do much or most of the brain’s learning and remembering: the amygdala is the specialist for emotional matters. If the amygdala is severed from the reset of the brain, the result is a striking inability to gauge the emotional significance of event; this condition is sometimes called ‘affective blindness’… The amygdala acts as a storehouse of emotional memory, and thus of significance itself; life without the amygdala is a life stripped of persona meanings… The workings of the amygdala and its interplay with the neocortex are at the heart of emotional intelligence.”

“…When impulsive feelings overrides the rational, the newly discovered role for the amygdala is pivotal. Incoming signals from the sense let the amygdala scan every experience for trouble. This puts the amygdala in a powerful post in mental life, something like a psychological sentinel, challenging every situation, every perception, with but one kind of question in mind, the most primitive: ‘is this something I hate? That hurts me? Something I fear?’ If so – if the moment at hand somehow draws a ‘Yes’ – the amygdala reacts instantaneously, like a neural tripwire, telegraphing a message of crisis to all parts of the brain… The amygdala’s extensive web of neural connections allows it, during an emotional emergency, to capture and drive much of the rest of the brain – including the rational mind.”

“The conventional view in neuroscience (prior to LeDoux’s research) had been that the eye, ear and other sensory organs transmit signals to the thalamus, and from there to the sensory processing areas of the neocortex. , where the signals are put together into objects as we perceive them. The signals are sorted for meanings so that the brain recognizes what each object is and what its presence means. From the neocortex, the old theory held, that the signals are then sent to the limbic brain, and from there the appropriate response radiates out through the brain and the rest of the body. That is the way it works much, or most of the time – but LeDoux discovered a smaller bundle of neurons that leads directly from the thalamus to the amygdala, in addition to those going through the larger path of neurons to the cortex. This smaller and shorter pathway – something like a neural back alley – allows the amygdala to receive some direct inputs from the sense and start a response before they are fully registered by the neocortex… The amygdala can have us spring to action with the slightly slower – but more fully informed – neocortex unfolds its more refined plan for reaction.“

“While the hippocampus remembers the dry facts, the amygdala retains the emotional flavor that goes with those facts… The more intense the amygdala arousal, the stronger the imprint… This means that, in effect, the brain has two memory systems, one for ordinary facts and one for emotionally charged ones. A special system for emotional memories makes excellent sense in evolution, of course, ensuring that animals would have particularly vivid memories of what threatens or pleases them. But emotional memories can be faulty guides to the present. One drawback of such neural alarms is that the urgent message the amygdala sens is sometimes, if not often, out of date – especially in the fluid social world we humans inhabit. As the repository for emotional memory, the amygdala scans experience, comparing what is happening now with what happened in the past. Its method of comparison is associative: when one key element of a present situation is similar to the past, it can call it a ‘match’ – which is why this circuit is sloppy: it acts before there is full confirmation.”

“One reason we can be so baffled by our emotional outbursts, then, is that they often date from a time early in our lives when things were bewildering and we did not yet have words for comprehending events. We may have the chaotic feelings, but not the words for the memories that formed them.”

“While the amygdala is at work in priming an anxious, impulsive reaction, another part of the emotional brain allows for a more fitting, corrective response. The brain’s damper switch for the amygdala’s surges appears to lie at the other end of a major circuit to the neocortex, in the prefrontal lobes just behind the forehead.”

“When an emotions trigger, within moments the prefrontal lobes perform what amounts to a risk/benefit ratio of myriad possible reactions, and bet that one of them is best. For animals, when to attack, when to run. And for us humans… when to attack, when to run – and also when to placate, persuade, seek sympathy, stonewall, provoke guilt, whine, put on a facade of bravado, be contemptuous – and so on, through the whole repertoire of emotional wiles.”

“The neocortical response is slower in brain time than the hijack mechanism because it involves more circuitry. It can also be more judicious and considered, since more thought precedes feeling. When we register a loss and become sad, or feel happy after a triumph, or mull over something someone has said or done and then get hurt or angry, the neocortex is at work.”

“Emotional hijackings presumably involve two dynamics: triggering of the amygdala and a failure to activate the neocortical processes that usually keep the emotional response in balance – or a recruitment of the neocortical zones to the emotional urgency. At this moment the rational mind is swamped by the emotional.”

“Take the power of emotions to disrupt thinking itself. Neuroscientists use the term ‘working memory’ for the capacity of attention that holds in mind the facts essential for completing a given task or problem, whether it be the ideal features one seeks in a house while touring several prospects, or the elements of a reasoning problem on a test. The prefrontal cortex is the brain region responsible for working memory. But circuits from the limbic brain to the prefrontal lobes mean that the signals of strong emotion – anxiety, anger and the like – can create neural static, sabotaging the ability of the prefrontal lobe to maintain working memory. That is why when we are emotionally upset we say we ‘just can’t think straight’ – and why continual emotional distress can create deficits in a child’s intellectual abilities, crippling the capacity to learn.”

Note: And to cripple one’s ability to do their job well if they are in a state of constant fear or anxiety or stress at work.

“In one study, primary school boys who had above-average IQ scores but nevertheless were doing poorly in school were found via neuropsychological tests to have impaired frontal cortex functioning. They also were impulsive and anxious, often disruptive and in trouble – suggesting faulty prefrontal control over their limbic urges. Despite their intellectual potential, these are the children at highest risk for problems like academic failure, alcoholism, and criminality – not because their intellect is deficient, but because their control over their emotional life is impaired. “

“In a sense we have two brains, two minds – and two different kinds of intelligence: rational and emotional. How we do in life is determined by both – it is not just IQ, but emotional intelligence that matters.

Posted on

PID Position Control Tuning With a Robot

A while back when I was working on tuning the control loops for the expressive degrees of freedom of our robot I thought I’d make a video showing a more fun example of classical controls tuning when you don’t know your system than what I typically see in classrooms.

For this, I’m using a DC motor as an actuator, hall effect quadrature encoder for position sensing, ROS with rqt_plot to visualize the position input and output, dynamic reconfigure to change the control gains, and have a few python scripts to generate and publish reference trajectories.

Step Response

I begin by just publishing a reference step function between 0 and 600 encoder ticks and start twiddling control gains! I’m doing this as quickly as I can to illustrate that you can tune things without actually looking at time domain step response parameters like rise time, settling time, etc.

PID Tuning Via Step Response

Note: at 3:27 I mispoke and say “proportional term” when I actually meant “integral term”.

Frequency Response

After first tuning my gains via visual inspection from a step response, I move on to a frequency response to illustrate what tracking of a varying frequency sinusoidal position input looks like.

The basic thing that I’m trying to convey to students is that the rise time and overshoot from our step response can in a way be translated to our frequency response. When the inverse of the frequency is faster than our rise time, we begin to see that the magnitude of our frequency response decreases. When the inverse of the frequency is at about the time that we see an overshoot in step response we begin to see the beginning of a resonant peak in our frequency response.

Frequency Response

Putting It All Together

I’m just kidding, this is me goofing off, but system dynamics and controls don’t need be dry and boring! Ignore the commentary, we were delirious from working long hours.