Some headway is being made on the mix of inductive learning and reasoning from information gained form the web. As an example please see the Read the Web project at CMU that uses NELL - http://rtw.ml.cmu.edu/rtw/.

Here is a nice overview article to a program of research conducted by Les Valiant of Harvard - http://people.seas.harvard.edu/~valiant/fsttcs08.pdf

Hello Daniel, Browsed through the knowledge infusion paradigms of NELL. Great.

Has NELL been able to connect the individual concepts into a new idea which was not yet parsed in the past?

Yes, it has. For a good overveiw see Tom Mitchell's Turing Awards talk at http://www.youtube.com/watch?v=51q2IajH94A

The mind (or brain) models the external world in some sense. This includes ontological models of the physical world, as well as the linguistic, social, ethical and cultural environment.

In fact, children do not in general know whether what they are learning is universal physical law, natural science fact, linguistic convention or cultural norm (see the writings of Piaget and work in Psycholinguistics). Learning about gravity, solid objects and hard knocks, the (positive and negative) information for learning comes from the physical world. Learning about fruit is a mix about learning about properties that mix the physical (Galileo and Newton could use apples as solid objects to learn about gravity), with the chemical and biological (what an apple or orange tastes like, that it is good to eat, that it is juicy), with the cultural (that we eat them in specific ways, that we can juice them in specific ways, that we eat them differently from each other or a grapefruit), with the linguistic (what words and grammar we use in association with them). A famous anecdote has a teacher showing a class of children a grapefruit which they recognize from colour and size and other physical characteristics as a grapefruit. But then she starts peeling, segmenting and eating it like an orange, and they change their minds and say it must be an orange.

Can we measure the effect of social context on a cognitive computer (whether an Artificial Intelligence or a Human Intelligence)? Yes, indeed we can... and it is an important source of information about how intelligence/cognition works. In fact, we always do, even though we may not realize it!! Every time we hear a sentence we interpret it in the physical, cultural, social and linguistic context. Here are some sentences that I have deliberately pseudophoneticized slightly to make more like hearing the speech - but in fact there will be prosodic context as well as (prior, current and/or subsequent or consequent) physical, social, cultural and linguistic context to help disambiguate the meanings.

Imgonnaworknow.

Timeflieslikeanarrow.

Imgonnarekanispeech.

Thehorseracedpastthebarnfell.

Thegreenscombinedwiththelibsvotedno.

The first one is understood differently depending on whether your wife knows its Saturday and you are not going in to work, or Friday and your are, or it is your teacher you are talking to, who is saying you are failing and you should drop out of the course.

The second is a famous artificial sentence, and could be about a particular kind of fly, timing flies, or time.

The third is also a famous artificial example that is understood differently depending on whether you are speech scientists talking about speech recognition, or hooligans talking about wrecking a nice beach.

The last two have the same grammatical forms and garden path - the second of the pair will give clues for understanding the first - it hinges on whether we interpret the -ed word as an active verb or a passive participle. The final verb phrase might be better replaced with a prepositional phrase (at the event on Friday, on the issue of abortion) so it can revert from participle to verb (but it still seems to require context to be a good sentence).

So these examples also tell you how. You test for the effect of any kind of context by devising experiments where the default or alternate understanding, or lack of understanding, is modified if the context is taken into account.

P.S. I'd like to hear more about your neurosynaptic machine interface. Is this some kind of (invasive) Brain Computer Interface, or some kind of Artificial Neural Network based Human Computer Interface? In either case, the question of human factors testing involving physical, linguistic and social context is relevant. In fact I am working on all of the above!

Hello David, great thoughts and examples.

I would like to know how can we meausre the relation between social context and human or machine cognition. Please find the details about IBM innovations on neurosynaptic machine interface below:

http://www.research.ibm.com/cognitive-computing/neurosynaptic-chips.shtml