Morning of Tuesday, June 2, 1998:
Some example questions for the N/X
What contributes to the understanding of the visual screen and what takes away from it?
People have a mindset as to how important something is based on its colour and size. For example, there will be a problem if you code red and big as "not important". Humans have intrinsic mappings in their psychology, and we have to match that. Red and big must mean "important".
Humans assume that the bigger and brighter things are, the more important they are. If we are not careful, the user may focus on an unimportant item on the display.
This is the notion of perspective, i.e., depth.
There are things that are individualised as well. We can do a lot to avoid very obvious errors, but we canāt make sure we have it right for every user.
The individual user part should be left to commerce. This makes it flexible enough for the individual user.
I disagree. We are not talking about an individual user who has a lot of time. We are talking about a commander who must use the information on the screen to make a decision.
What contributes to understanding is user driven visualisation.
Why canāt you let the user pick their colours?
You canāt do that·
I think that it is important to realise that there are two different types of understanding. It is important to keep track as to whether the understanding can be done perceptually or if it has to be done in a much more cumbersome manner (i.e., logic).
With colour, there are cultural biases. Also, you must maintain continuity from one screen to another. You donāt want to give the user freedom on colour as it will cause many problems.
We did some experiments with the use of colour in sonar. If you colour code intensity, you end up with colour speckles. Colour may be useful in alerting, i.e., you can highlight a contextual area where there might be a target.
When using colour, you need consistency and relevancy.
If all of the data is intensity data, all you need is black an white.
This issue is one that I thought would be sorted out long ago. I think we should look in reference documents and just use that information. If someone disagrees when they are designing their own display, then fine.
Some of colour research may have got some things wrong ö I think that it should be revisited.
Colour is very important. I created a colour display for temperature, and it was very effective.
Humans associate red with alert. There is a natural expectation and it is easier than training the user. We also must look at yellow.
Colour seems to be culture specific.
We have been talking about perception but we have forgotten about the physical systems. We have to develop systems for people with vision problems ö there are differences between peopleās abilities to distinguish between different colour combinations. For instance, people who are colour blind canāt distinguish certain colours.
There have been many books written on colours.
Haim Lefkowitz wrote a book on colour entitled Colour on the Computer. It is on the web.
There is a huge body of work on colour. Within the body of work there is discussion about choice. It contributes to a user understanding when the colours are relevant.
The key is that there must be some distilled fundamental things we must understand about colour, i.e., personal preferences. We should document some works about the impact of colour in visualisation. We should see if there are some real derivatives, i.e., universal relations to colour.
Haim wrote about which colours are easy to discriminate. His work must be referenced. We also need to reference other works on colour displays.
Tonyās model is a concrete model. Sharonās is an abstract model based on sound. There are extreme differences in the underlying models that they used.
Zack Jacobson on the White Board:
Use of COLOUR for visualisation
Colour à alerting
(already automatically processed)
User à choice à ABSTRACT
| confirmatory |
| disconfirm |
of false positives
Zack Jacobson (referring to the white board):
This is intended as a model paradigm.
The issue ties into the intrinsic/non-intrinsic mappings in displays. Mapping tanks is intrinsic. When you are mapping sound, it is abstract. There is an issues as to how it should be done.
When you start to work on a new problem, you look for existing viewing paradigms and you use those to ease the visualisation task. Visualisation should allow a person to visualise and understand a structure that they originally did not understand.
During the discussion on sonar, there was discussion about false negatives and false positives. The question is, is visualisation being used to confirm or reject our hypothesis, i.e., are we trying to accept this or reject this as truth. I think that people have differing views on this.
Where does this fit into your four tasks, Martin?
Logically, there is a relation. They seem to be done differently in different parts of the brain, so there are different requirements for testing. Fitting these two aspects together will be very important.
We do different things when we are confirming or disconfirming something.
It doesnāt bother me that the alerting system has false alarms. But if you have to go through some over-manipulation to examine them, it becomes a problem.
Is a disconfirmation an alerting?
The purpose of an alert is to indicate that in a large uninteresting field, there is something of interest.
A problem that occurs when the ratio of satisfactory alerts is too low. If it is, you are not going to recognise the confirmations when they occur.
Some alerts are easily dismissed, but there can be a problem. For instance, in an aircraft, pilots disconnect warning systems due to false alarms. They are not alerted when something occurs, resulting in a crash.
How do you distinguish between security and insecurity?
There is a value vector involved.
You canāt afford not to detect a submarine.
You can use the area of game theory (accuracy).
It is different if you miss the submarine and no one ever saw it versus if the alarm saw it and you never did.
From my experience on a submarine, you never stop jumping when you hear the alarm.
There is a sense of a loop here that ties into collateral information. It comes back to "I want more data".
Is there some sort of system that prioritises the alerts after they go off?
That is "search".