Putting Context in Context

Not long ago, Dr. Robert L. Brennan, a professor and friend of mine, identified what he called the "context" of "context effects." He later published the research in Applied Measurement in Education. Ever since then, I have been intrigued with the notion that it is not the context effects themselves that play havoc with measurement, but rather it is the changing context in which the effects occur that seems to cause the problem.

Recently, a friend forwarded to me an email that again peaked my interest in the role of context in learning and assessment. Here is part of the email:

Can you raed tihs?

i cdnuolt blveiee taht I cluod aulaclty uesdnatnrd waht I was rdanieg. The phaonmneal pweor of the hmuan mnid, aoccdrnig to a rscheearch at Cmabrigde Uinervtisy, it dseno't mtaetr in waht oerdr the ltteres in a wrod are, the olny iproamtnt tihng is taht the frsit and lsat ltteer be in the rghit pclae.

I am told that this is the "trick" to speed reading—namely chunking the information at a level of aggregation higher than the letters' words. This phenomenon—being able to read such jiberish—makes sense when you stop and think about it. Context provides much of the things we claim to be "understanding."

Take for example the following sentence segments: "You have a hot car." "It is very hot today." Technically, you can't tell what the definition of hot is without the context. As such, knowing the primary definition of the word "hot" will do very little, if anything, in helping you understand the meaning behind these sentences. This dilema is one of the reasons that reading is difficult to teach. It is also what causes most automated/computerized "readers" to fail.

However, Latent Semantic Analysis (LSA) is a promising, if not new way to use context to allow the computer to actually infer meaning from sentences or other snipits of information in text. As such, the computerized "readers" used in engines like the Intelligent Essay Assessor (IEA) provided by Pearson Knowledge Technologies (PKT) are likely to finally allow the computer to infer meaning from text (student generated or otherwise) at accuracy rates acceptable to the measurement community.

But don't take my word for it, check out the research yourself : Reliability and Validity of the KAT™ Engine.