ENP162->Blog 3

about information theory, signal detection

Signal Detection

Here is how you could understand the signal detection in one picture.

How to Use Signal Detection Theory

There's a lot of signal detection examples online.

First of all, it is a means to measure the ability to differentiate between information-bearing patterns and noise. It is also used when psychologists want to measure the way we make decisions under conditions of uncertainty, such as how we would perceive distances in foggy conditions or during eyewitness identification. There are great resources about the application and benefits:

“UTILIZING” SIGNAL DETECTION THEORY:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4304641/

APPLYING SIGNAL-DETECTION THEORY TO THE STUDY OF OBSERVER ACCURACY AND BIAS IN BEHAVIORAL ASSESSMENT

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2884345/

If We Focussing on False Alarm Specifically

False alarms have been talked about a lot. Why? It has the potential to divert emergency responders away from legitimate emergencies, which could ultimately lead to loss of life.

Here's one examples you might want to look at:

Signal Detection: Hits and False Alarms Examples

http://wise.cgu.edu/wise-tutorials/tutorial-signal-detection-theory/signal-detection-hits-and-false-alarms-examples/

So is there anything we can do to prevent false alarm?

Stay clam, here are the top 8 causes and what you can do to prevent them. Many thanks to those who summarized and write it down.(https://help.brinkshome.com/hc/en-us/articles/360002586512-How-to-stop-false-alarms-now)

1. Human Error

Residential

• Use of incorrect keypad codes

• Failure to train other authorized users (i.e., sitters, relatives, children, visitors, repairmen, dog walkers, etc.)

• Fail to cancel alarm with monitoring facility

Commercial

• Use of incorrect keypad codes

• Failure to train other authorized users (i.e., employees, janitors, delivery personnel, cleaning crews, etc.)

• Failure to cancel alarm with monitoring facility

• Failure to notify monitoring facility of unscheduled openings or closings (for businesses using set schedules)

• Failure to update authorized personnel list with monitoring facility

How to prevent it: Make sure anyone who needs to use your security system has been properly trained, respond to alarms promptly, and proactively provide all updates to our Customer service team.

2. Accidental Activations (caused by keyfob panics)

• Keyfobs panic button becoming lodged or stuck

• Children playing with keyfob remote

How to prevent it: Be sure to keep your keyfob in a safe place when not in use, refrain from leaving in pockets. 

3. Smoke Detectors

• Placing smoke detectors near bathrooms, kitchens, or in areas where steam or dust can accumulate in the detector's chamber

• Failure to replace batteries in smoke detectors

How to prevent it: Clean smoke detectors with a vacuum cleaner or a can of compressed air periodically to get rid of built-up dust. In addition, cover up smoke detectors temporarily during construction or projects in the home.

4. Improper arming and disarming of the system

• Failure to arm and exit the home by the exit delay timeframe

• Failure to disarm the system within the entry delay timeframe

How to prevent it: Have your user codes for the system memorized and be aware of your entry and exit delay timeframes. If you notice you're frequently exceeding the timeframe (and setting off the alarm) contact Customer Technical Support to update your settings. These alarms can also be prevented by entering your home with at least one hand free to properly disarm your system.

5. Pets, rodents, and insects

• Failure to install pet-friendly motion detectors with adjusted settings

• Rodents and/or insects near sensors

How to prevent it: Purchase an alarm system that is tolerant of pets (pet immune). Spray insect repellent around sensors and detectors twice a year. Contact a professional exterminator if rodents, insects, or other pests are a repeated problem.

6. Lack of maintenance (dead batteries)

• Failure to replace batteries in equipment (Motion Detectors, Smoke Detectors)

How to prevent it: Service and maintain the system (including replacing batteries) properly.

7. Unsecured windows and doors

• Failure secure doors and windows before turning on the alarm

• Loose doors or windows (door or window jiggles or is difficult to close)

How to prevent it: Confirm doors and windows are shut securely and locked. 

8. Power problems

• An inconsistent/unreliable power source

How to prevent it: Ensure your transformer (power source) is secured to a working outlet. Multiple power outages can cause a weak backup battery which can cause an alarm.

Thoughts

I am interested in how we could use the single formula to measure information and how we could use the information theory into human factors engineering. First thing I did is google information theory and Shannon entropy.

Look at this beautiful formula:

With this formula, we could calculate the average members of bits. The example they used is about coin flips. There're four possible outcomes(p1=9/16,p2=3/16,p3=3/16,p4=1/16). We could calculate the average bits per sequence(also known as Shannon entropy). We got entropy equals to 1.6875 (H(x)=−∑xP(x)logP(x) )

H() is the convention/notation used to represent Shannon entropy; this is the value that is our average number of bits.

X represents the set of all possible outcomes; in our example this is the four possible sequences.

For the entropy, here the dice example, the chart is clear and easy to understand. Smaller entropy, smaller "surprise value" (or " degree of unexpectedness" or " uncertainty about the outcome").

This brings me to thinking. As a designer, we normally want the entropy to be smaller so that our users could have a "non-surperise situation". How could we reduce the entropy?

I designed four situations for me to solve my own question:

let's assume there're event a, b, c, d, probability are:

1) A=b=c=½

2) A=5/16, b=5/16, c=6/16 (smaller difference)

3) A=7/8, b=1/16, c=1/16 (bigger difference)

4) A=½, b=½

Based on the formula,

1) H(x)=1.0986

2) H(x)=1.0948

3) H(x)=0.4634

4) H(x)=1

Look at the result, it's obvious that

1) bigger (probability) difference between the three events, smaller entropy.

2) fewer events, smaller entropy

This result directly points to two famous theory:

1) Fitts's law--more chances, less confusion.

Increase the probability to decrease the certainty. for example, a large size button increases the probability to be chosen, so that user feel less surprise when they need to click. because the button is big, and they might feel they are supposed to click that button

2) Hick-Hyman law--fewer options, less confusion.

Those two laws are different ways to reduce entropy, I think it is curial for human factors engineers to understand why and how it helps.

Anyway, thank you for reading this! If you were confused about information theory, I hope my blog helps.