Human Bias - Collaborative Identification

At it's most basic, birding is a lone pursuit.  Identification begins from the moment the first sights and sounds of a bird engage the human senses.  But, shared birding experiences are often better.  Collaboration in anything brings rewards, be it democracy, project and committee work of bird identification.  In the past, unless birders were out in the field together collaboration in bird identification tended to involve discussion around some field notes and possibly some sketches... photos if you were lucky.  The internet and the digital era have opened up a whole new avenue for discussion and there are numerous forums and chat rooms devoted to birds including bird identification.  But what frustrates many people about social media is the lack of decision making power, consensus and action.  Very often discussions allow ideas to be aired but that is where things fizzle out.

Collaboration in decision making online just got a whole lot more interesting with the arrival of Loomio from a collaboration of organisations in Wellington, New Zealand.  While this brilliant tool has been seized initially mainly by social and democratic movements, in time this may become a very everyday mechanism for online decision making.  Perhaps just as commonplace as social media itself.  In time could this technology even mark an end to the classic rarity assessment process?  Consider an all-encompassing website that compiles records, encourages consensus on bird identification and generates and maintains the official list for a region or nation.

Of course there are potential pitfalls, including the potential for populist opinion to dictate an identification.  Bird identification is not a democratic process.  A bird only has one identity.  The goal therefore of collaborative identification is to reach the correct conclusion, not the populist one.

So what safeguards need to be built in to such a process?

Rarity Assessment Panels
The primary goal of a rarity assessment panel is to maintain the integrity of an official regional or national list.  For anyone who has participated in a rarity assessment there are multiple challenges including understanding the criteria for identifying the subject, the abilities to objectively and correctly assess the evidence, the ability to think laterally when needed and perhaps a certain level of scepticism and an eye towards biased or even embellished note-taking.  Typically records are assessed by individual voting members and records may go through recirculation more than once before if necessary being discussed more collaboratively to reach a consensus.  Expert opinions are often sought from outside the committee in order to ensure a correct outcome.

I often perceive birders as among the most diverse group of people one could imagine.  Birders the world over share the exact same love and passion for birds, which is what makes it such a universal interest.  What surprised me most about my time on the Irish rarity assessment panel (the IRBC) was the singular focus by everyone that I served with on maintaining the integrity of the Irish list.  All voting members down through the years seem to have had an inherent hunger to reach an accurate conclusion as to the identity of each bird and also to validate each record.  Only when those objectives could not be reached satisfactorily was there disquiet among the committee members.

The more I have thought about these fundamental aspects of rarity assessment the more I have come to realise that these form part of what drives all birders.  Birders want to get their identifications right and really hate it when that bird just gets away or is simply too difficult to identify.  This is part of what drives human cognition generally.  We strive to categorise and label everything and this overwhelming drive can lead to a number of human cognitive biases.

Bird of Unknown Origin
Most twitchers hate to hear a bird labelled thus.  Very often decision making for a rarity assessment panel extends beyond the mere acceptance of an identification to an attempt to divine its origin.  Many potential vagrants happen to be kept in captivity and national and regional lists tend to exclude records involving birds which are suspected to have escaped from captivity.  The true origin of some birds can be a matter for considerable and heated debate.  In many cases rarities assessment panels tend to take a more conservative view of these things than many within the general birding community, particularly those with an interest in regional or national listing.  For an online collaborative assessment system to deal with this conundrum would require some thought but I still think it could be overcome.  One way of doing so would be to establish criteria for acceptance of potential vagrants and to maintain an ongoing review of such criteria as circumstances change and patterns emerge.  This at least creates a framework for general agreement.  But the reality is no individual or group can ever divine the true provenance of a bird in the wild without a ring or other marker to pinpoint it's true origin.

Collective Bias
Of the many forms of cognitive bias group think may be among the most pervasive.  I like the simple and effective pie-chart graphic used by Loomio to illustrate consensus building.  This certainly lends itself to the premise of democracy that the majority rules okay.  I wonder however at the same time if this graphic actually encourages group think.  After all if I join a discussion where already my opinions are at odds with the majority I may feel a natural pressure to conform with the group.  Of course then there are the rebels who are more at home in the minority percentile and for whom conformance and agreement with the group is an anathema.  There are many birders who simply don't accept the peer review of their sightings.  No system however democratic will invite 100% take up.

There will always be leaders and followers.  Collaborative decision making in theory allows for all voices to be heard but some voices will always be louder or more persuasive than others.  Even among seemly democratic structures cliques and factions form.  Meanwhile lobbyists and bullies also have a disproportionate influence on outcomes in many cases.

I have often felt that the best way to approach any complex question is to exclude myself from the noise of debate and first try to resolve the answer for myself.  I believe collaborative discussion is the best way to reach a decision but I think it is really important for everyone to have clarity of the subject before engagement.  I believe that any online tool that might in time be used for assessment and identification should have at it's heart a measure of built-in protection to prevent collective and other forms of cognitive bias from influencing the decision-making process.  Examples might include blind surveying, anonymous contributions and statistical process controls etc.

Conclusions
Bird identification for many is a very personal pursuit.  Many of us learnt to identify birds on our own through trial and error.  And, many prefer to continue to work that way.  Collaborative identification occurs all the time in the field, through online chat rooms and email groups, and within the deliberations of rarity assessment panels.  In the near future we may begin the see the replacement of formal rarity assessment panels with a more collaborative online assessment process.  We need to be mindful however that bird identification is not a democratic process - there can after all only be one correct outcome.  But we also need to learn to accept that some birds are best left unidentified and that the true provenance of most birds can never really be known.

Birds and Light - Foliage Canopy Edge

In the posting Under Foliage Canopy I discussed lighting impacts when studying and photographing birds under a foliage canopy.  Late spring and early summer in temperate forest zones in the Northern Hemisphere is characterised by a vivid green light produced by new foliage growth as illustrated below.  By late summer forests in temperate areas begin to take on some of the characteristics of forests of the tropics.  The vivid green light which baths these forests in early summer gives way to a gloomier light as the canopy closes and as leaves harden and becomes less translucent.  This trend persists into the autumn until deciduous leaves begin to discolour and fall, reopening the canopy once again.  

In this posting I am specifically looking at photography along that boundary, looking into the canopy from outside, where birds move in and out through the foliage.  And, just for clarity, here I am looking solely at mature canopy cover rather than new growth.  So, whereas in these conditions green transmitted light can still certainly play a role, the real lighting challenge on the canopy edge tends to be changing light intensity and not so much the colour of the light.  On a dull day such as today in Ireland it can be very striking just how dark and forbidding foliage canopies can appear.  Due to the dense growth of late summer the light intensity seems to start falling off strongly just inside the boundary of the outermost leaves and branches.  A lot of birding actually takes place on the edge of the foliage canopy, whether it is along a forest path, or along a hedgerow or in a park or garden.  Obviously many birds like to stay close to or within cover so we are frequently dealing with this challenging, changing light.

Metering and Exposure Control

I have covered camera exposure and light metering in depth HERE.  Metering subjects that are moving along and through the edge of a canopy can be extremely challenging due to the nature of the metering process and the fact that the lighting on the subject may be constantly changing.  Sometimes it pays to select a wide metering scope such as Evaluative Metering based on the canopy and subject rather than Spot Metering on the subject alone in order to keep the exposure range narrow, then use exposure bracketing to generate fixed exposure increments in the hopes of capturing a few decent shots.  The alternative is to use Spot Metering and try to carefully track the subject closely and have the camera adjust exposure accordingly as the lighting changes.  But if we are not metering off suitable surfaces we end up with poor exposures as described below.

HDRI and Tone Mapped Images

With bracketed exposures we have the potential to merge exposures to generate High Dynamic Range Images (HDRI).  I have explored HDRI HERE and as the posting demonstrates HDRI is only really practical if the subject is completely motionless and a tripod is used.  HDRI type images can also be produced manually from a single RAW image file, referred to as Tone Mapping.  This may be a slightly more practical solution.  This is also discussed under the HDRI posting and demonstrated with the European Robin (Erithacus rubecula) image below.

Ghostly Subjects

I am currently trying to get to grips with a phenomenon that has perplexed me for years.  Very often when I have observed and photographed a very pale, ghostly subject such as a pale warbler on the edge of a foliage canopy I have been very disappointed with how seemingly unrepresentative the images of it have been.  Quite often the images appear too dark to represent the subject, suggesting that the images may have been underexposed.  Then if I happen to have captured images that look more representative of the bird in life, the images to me often appear overexposed and not quite accurate.  I am sometimes left wondering if in fact these birds are not quite as pale in life as they appear to be and if it could all simply be a Brightness Illusion as discussed HERE.

Here is another example of a very pale bird which for me never quite photographed correctly.  This Marsh Warbler Acrocephalus palustris proved to be a difficult subject to photograph as it only appeared briefly from within the brambles from time to time.  Like the Aberrant Willow Warbler Phylloscopus trochilus example above it was digiscoped from some distance away, which only added to the difficulty of trying to obtain a proper exposure.

As I discussed elsewhere under the posting Grey Scales and Gulls, without having a frame of reference such as a grey card to guide a correct exposure the actual correction for the individual tonal range of a subject can be a matter of pure guesswork.

Exposure and RAW Tone Mapping Experiments
Using a lux meter I carried out a simple experiment to test how light drops off as one penetrates the foliage canopy.  I selected a single example of three common trees and started with a reading out in the open and a number of readings at arbitrary distances underneath the canopy but without disturbing the leaves.  This simple experiment reveals just how quickly and effectively a mature canopy absorbs the available light.  In the case of the Leylandii Cyprus 99.9% of the light was absorbed within a half a meter of the canopy edge.  Both the Alder, Alnus glutinosa and the Common Hawthorn, Crataegus monogyna absorbed 80% of the light before I reached a point where I couldn't proceed any further without disturbing the canopy.

Next, to demonstrate how slight changes in exposure and distance from the foliage canopy edge can alter the appearance of our subject I placed a number of identical white targets on the surface of leaves of an Alder, from the very edge of the canopy to approx. 20cm and 40cm inside the edge respectively.  The day was overcast and dull when all these images were taken.  A drop of a few hundred lux in this case within a few centimetres of the canopy edge clearly had a fairly dramatic impact on our subject.

I took a bracketed exposure involving normal, 2 stops over and 2 stops underexposed.  By shooting in RAW we can maximise the quality of our images as discussed HERE.  I selected the best of the three bracketed exposures and opened up the RAW version of the image.  I then set about Tone Mapping the image by eliminating clipping at both ends of the histogram and trying to bring out the maximum tonal range on all three targets.  The results show a significant improvement on the earlier results.  This may be the best way to obtain consistent results when photographing birds on the foliage canopy edge.  The challenge with all photography is to obtain an image that represents the tonal range of our subject.  Bracketed exposures, coupled with tone mapping gives us a good chance of achieving that goal. 

Human Bias - Brightness Illusions

Before reading this post please visit Lottolab Studio and have a look at a number of their brightness illusions.

I have already referred elsewhere to luminance and tonal illusions HERE, HERE and HERE among other postings.  I have discussed for instance the Checker shadow and Chubb Illusions.  These both depict lighting illusions (shadows and tonal gradients) in greyscale.  The work of Beau Lotto opened my eyes to related illusions concerning colour hue and tone.  It seems that very similar mechanisms are involved whether an illusion entails simple brightness or colour tones.

What is a colour?

Colour is an entirely human construct.  We are not born with an innate understanding of colour.  Colour is learnt, often quite slowly it seems.  From a basic vivid child's colour pallet we can learn to distinguish and possibly name many thousands of distinct colour hue, saturation and luminance tones.  For survival we have an innate attraction to certain colours in nature and are better able to see some colours more than others.  Meanwhile, physiologically our green colour detectors (cones) outnumber red and blue cones two to one so our colour vision is somewhat biased.

The Eye Versus The Camera

Unlike the camera, the human visual system is truly dynamic.  Perhaps the light sensitive cells of the eyes record light just as simply as the photosites of digital camera sensors but the way our brain then processes this information is far more complex and nuanced than that of the camera processor.  Most in-camera processing involved adjustment of all pixels (global adjustment), though there may be some finer tuning at the broad tonal level (highlights versus mid-tones versus shadows).  The Lottolab examples point to a far more targeted image processing by the brain, involving small (local) parts of an image.

Given this level of complexity human vision is perhaps less helpful for forensic analysis of images than we would like to think, particularly where there is significant individual variation between observers.  The amazing Dress Viral Phenomenon of early 2015 demonstrated just how polarised individual perception of colour and lighting can be.  We also know that a range of other Human Biases can come into play when an observer is requested to describe or recall what they see.

Optical Illusions - Fooled or Not Fooled

I have found myself a little confused by the relevance of some of Lotto's Illusions.  Among the various examples are three shown below.  Clearly these should evoke the same illusion as the Checker Shadow (particularly the top right example).  But in these cases I have failed to see the illusion.  To my eyes the tones involved appear the same whether the images are masked or unmasked.  That said I suspect if I come back to these again in a week or two perhaps I wont be as prepared and the illusion might 'kick in' then.

Where my eyes clearly do deceive me every time however are in Edward H. Adelson's Checker Shadow example and the very similar coloured cube tonal illusions from Beau Lotto.  Both are simply mind-blowing illusions.  Very often when I look at these I experience a transition over a second or two as my eyes, or rather my brain adjusts to the reality that these tiles are in fact identical in tone and therein lies a fundamental truth about human vision.  Our vision is entirely dynamic, designed to cope with an ever-changing world and to quickly adapt both globally and locally to almost any lighting conditions.

Computer Generated Images Versus Real World Images

It could be suggested that computer generated optical illusions are not so easily replicated in the more complex lighting environment of the real world.  It is worth watching Beau Lotto's lectures online to see how he masterfully brings illusions to life using simple colour cards, lights and filters.  So clearly these illusions don't just exist on the computer screen.  However, if these illusions do frequently occur in nature, why don't we experience the kind of 'eureka moment' in everyday nature that we do when the illusions shown above are unveiled to us.  Surely, every once in a while we should find that a certain colour tone on a bird has weirdly morphed before our eyes, revealing something of the illusory nature of our vision systems.  Well perhaps occasionally this does happen.  It is often reported that exceptionally pale birds never photograph properly.  Could it be that our perception of these birds as being exceptionally pale is merely an illusion and that the camera is provides a more accurate representation?  This is something which I have been trying to investigate for some time and hopefully I will be able to write about it in due course.

If these illusions are to be found all the time in nature they must be masterfully hidden for a lot of the time.  With this in mind I set about to find some of these illusions hidden within bird images.  For this experiment I have focused on images showing similar characteristics to the Checker Shadow Illusion.  I have selected birds in partial shade to try and uncover an illusion similar to the ones shown above.  Here is a flavor of what I have uncovered.

Gestalt - Simple 3D Modelling from 2D images

With the massive growth of processing power modern home computers and tablets have the capability to bring to our fingertips the kinds of tools we might have only imagined ten years ago.  I have been dipping my toe into 3D modelling in the last couple of weeks and already have come up with some nice surprises.  The first is a wonderful phone app called Seene which allows anyone with a smart phone to generate simple 3D models in a matter of seconds.  Here is a Common (Black-billed) Magpie Pica pica tail feather placed on a wooden board.  Simply by moving the mouse over the image you can gain an impression of the 3D depth of the image.  For more click on the Seene logo at the bottom of the image.

A really nice little tool and certainly useful I think for capturing flowers, insects, such as moths and perhaps birds in the hand.  But not a very practical solution for birders trying to model birds in the field.  Incidentally, the Seene type images are pretty much what I had in mind for 3D image capture at the end of my posting on measurements from photographs, HERE.

But is there a better solution out there?  Well I started looking at some open source software and found a very nice offering in the form of a program called Visual SFM.  To start with I gathered a range of images of a Ring-billed Gull Larus delawarensis and cropped them in order to try and obtain a 3D model of the head and bill.

The first step involves comparing all these images for points which are common to two or more images.  Normally the intent with 3D modelling is that the person taking these images is moving around a stationary subject, but in this case I was stationary and the bird that was moving around.  So I was not entirely hopeful this method would work.  Nonetheless I proceeded to the next stage in the process.

Step 2 in the process involves generating a point cloud.  This is essentially a cloud of dots making up the surface of a 3D model.  The more dots the better the model.  To have achieved any dots at all was a surprise I'll be honest!

While it may not look like much here is the point cloud generated from the images above.  The next stage in the process involves creating a surface mesh from the available points and placing texture over the mesh.  Only at this point does the model begin to resemble the subject.

A second free software programme called MeshLab allows viewing of the end result.

I have to admit I am really shocked that this method produced any results.  What I was keenly interested in here was the facility to appreciate the bill to eye ratio and general proportions of the bill and head as discussed in an earlier posting HERE.  The model created certainly allows for some appreciation of these proportions.

I believe the major gaps in the data in this instances was down to the bright light.  It is recommended to capture images in diffuse light in order to minimize bright light and shadow.  I have no doubt that by fine-tuning this method better results are possible.

Here is a short video showing the 3D model in motion on MeshLab.  Further down is an instructive video demonstrating the various software elements and the true power of this method.

LINK

LINK