thank you for your answer,

i am interested in any kind of (as general as possible) attention metric from simple categories as "bored, interested, distracted", to attention types "focused attention, selective attention..." to scales from 0 to 10 as long as I can find some scientific basis for the specific metric.

I am aware of the investigation of pupil dilation as an indicator, yet this does not give me a mapping to an attention "score".

what do you mean by combined metric?

There is no single number for attention, because it is not a unitary construct (spatial attention, object-based attention, working memory allocation...).  Dual tasking )i.e., performance on a secondary task while doing a primary task of interest)  is sometimes used, but the measure would be specific to the secondary task.

thank you for your answer

i am well aware that there is not a single value which can cover all aspects of attention,

i am just trying to collect different approaches towards the qualitative description of attention

Are you familiar with the Connors Continuous Performance Test? It is a computerized attention challenge which runs for 14 minutes and presents the user with targets (white letters of the alphabet against a black background) and a non-target (the white letter x). The letters appear briefly on the screen at rates of 1, 2, and 4 second intervals and the user is directed to press the space bar each time a target is presented and to not press the space bar when a non-target appear. The software then scores the results by measuring reaction time, omission errors, commission errors, and so on. It is age and gender normed.

thank you for your contributions, 

@vincentkoenig, thank you for this hint, we will be emplying SMI products in the near future anyway so I will have a look at this concept

@stepehenmoores, I am familiar with the concept, and there are many more similar tests which analyse task performance as equivalent to attention, which will definitely be part of the overview of attention metrics I am currently pursuing. Yet, to apply these kinds of metrics, I need to formulate any behavior as dividable into quantifiable tasks to calculate performances, which is very limiting when aiming at non-laboratory general behavior analysis.

Another suggestion, although you may have considered it already, is the measurement of executive functions (EFs). EFs are thought to be top-down processes responsible for the focusing and maintenance of attention. The three core EFs are working memory, inhibitory control and task switching/cognitive flexibility. A suite of reliable and representative performance-based tasks (not self-report) capturing all three can be as short as 5 minutes. If you're unfamiliar with these and interested, I'd strongly recommend "Executive Functions", by Adele Diamond (Annual Review of Psychology, 2013).

@simon, thank you, I#M gonna have a look at it!

Hi Benedikt, I was at a CIKM workshop back in 2007 on contextualized attention. Unfortunately, I don't think it managed to gain momentum enough to arrange again. Nevertheless, perhaps you can find something useful in one of the papers or perhaps in their references. You can find the proceedings here: http://dl.acm.org/citation.cfm?id=1183604 

@lars thank you very much! I will look into this

For quantification you could look to physiological measurements, but I don't know whether your system supports that. Otherwise, as you have surmised from the literature, I think you will have to base your metrics on performance of a number of basic attention tasks.

you can use the english version of TEA test to measure attention (http://www.psytest.net/index.php?page=TAP-2-2&hl=fr_FR). it allows to measure many component of attention

Take a look at https://www.cisuc.uc.pt/publication/show/3723 also available in https://www.researchgate.net/publication/261265169_Arguments_for_a_Computational_Model_for_Forms_of_Selective_Attention_Based_on_Cognitive_and_Affective_Feelings?ev=prf_pub

Conference Paper Arguments for a Computational Model for Forms of Selective A...

Attention is drawn to things remembered or absence of background.  The latter is illustrated in modelling visual attention - http://www.ee.ucl.ac.uk/~fstentif/attention.pdf.

The same would probably apply to other forms of data.

We used a dual task approach to try and indicate attention allocation, with a simple reaction time task (responding to frequent but randomly spaced beeps by pressing a button) that ran concurrently with the main task of interest (walking down a street). We interpreted slow reaction times as suggesting attention had been diverted away from the reaction task towards something else, possibly something attention-grabbing or salient in the environment (we were using eye-tracking to see what people were looking at during the attention-diversion times).

http://lrt.sagepub.com/content/early/2014/02/11/1477153514522472

wow so many answers! thank you all in advance

@addiejohnson we are actually trying to use physiologial measurements, yet I still need tocome from a sensor input to which e.g. describes pupil dilation or heart rate to an expressive representation of attention or attention score.

@nicolascarvalho thank you, my problem is, that we do not want to put users in direct performance task situations, but want to perform a remote observation

@luismacedo thank you I will look into this

@fredstentiford, in my understanding, saliency describes which stimuli are likely to attract attention  and allow interpretation and identification of objects that may draw attention, yet this more concerns the control of attention than a quantifiable representation?

@jimuttley thank you, did you also apply any attention metric? or did you concetrate on the reaction times?

Hi Benedikt, for our purposes the reaction times were essentially our attention metric - longer reaction times suggested reduced attention towards this reaction task, and potentially greater attention towards something else.

I'm far from a specialist, but if it is visual attention you are interested in, you could check the Theory of Visual Attention by Bundesen (Bundesen, C. (1990). A theory of visual attention. Psychol Rev, 97(4), 523-547.). It is a mathematical, psychometric model that quantifies different aspects of visual attention (perceptual threshold, visual short term memory capacity, visual encoding rates). Its parameters can be fit from data. Hope this helps!

Selective attention can be observed in some quantity in  two types of behavior, both arising  when a person's observations don't meet his expectations ,which have been generated by a (mental) model.  Both types of behavior are aimed at reducing the discrepancy.  In the first  he looks for additional  observations to confirm the model.   In the second he constructs or recalls another model that better fits his observations.  That's a bare-bones version of Friston's free energy principle. A related idea developed in my chapter that appears in the book "Mind as Motion"

I attach a file that will serve you, I prepare myself in an attempt to explain the phenomenon. a greeting

given that attention is not equal awareness those scales are of no use for your tasks. Vigilance or measures of sleepiness may though provide you with a scale. More objectively might be to use the EEG, as alertness/attention towards a stimulus is signaled in the power spectrum i.e. beta waves - non-invasive and two electrodes on the scalp suffice (see: Jung et al. 1997 Estimating alertness from the EEG power spectrum)