Photos generally look older in black and white, but photos that look dated in color (due to photography tech) look newer in black and white where the photographic color artifacts aren't visible.
Eyeballing the data line they provide makes me think most people struggle to date photographs.
What surprised me is the magnitude of the effect in photos that were clearly taken after the advent of widespread mainstream colour photography. I'd have guessed the difference between the colour and non-colour groups would shrink considerably since colour would be much more a reflection of artistic choice than technological or resource constraint.
From the start I had a suspicion that they might be playing tricks with the image, so I tried to discount color (either the lack of, or sepia tones) and low resolution (I assumed that high resolution cannot be faked), and go by the clothing style as much as possible, and other items in the image (signs, use of horses). This made the large crowd at the Washington monument difficult to place.
I tried the same approach and the annoying thing about the mall photo is that it could be either MLK's march or the Million Man march in 1995. Without detail, you can't really tell unless you've seen the exact photo before. I picked 95 even though they showed me a black and white image.
I learned I don't really know much about clothing styles.
From the site:
>When you were dating the photos, you probably looked for context clues — what people were wearing, if there were any familiar buildings, and if you recognized any faces. You were probably also looking at color.
I would argue that the very nature of the site makes it more likely that most people were using context clues. I also wonder if color gives you more detail/clues and hence in most of the photos I was shown, the color guesses were closer to the actual year.
I would find the accuracy of the guesses (as opposed to the difference between color and B&W guesses) more informational, for a larger data set.
I did the same, while also using historical knowledge of events and things like architecture and relative spread of cultural or fashion clues. I would love to detail how I pegged basically all the images, but that is not something that can be discussed here for reasons I will not elaborate on.
It's actually an interesting point that is directly related to the subject research. How can you get an accurate picture when things cannot be discussed freely … you simply can't.
I encourage everyone to diversify your channels for for discussion and knowledge, be it distributed forums or even things like using some of the browsers and/or plugins that allow for open discussion on top of what is permitted by the heavy handed types (to put it it extremely lightly) among us.
I would like to know the effect of age (specifically if the photos were taken before we reached about 15 years old and really started to watch the outside world) on predictions.
The Einstein photo, for example... I knew he died in the 1950s, but I wasn't sure when.... but the audience was mostly Black, so that pushed it forward for me into the 1960s... without those clues, I would have been more wildly guessing.
I tried really hard to ignore the monochrome/color aspect... it would be really nice to know how far off I was on each after the fact.
When I finished it, I got a congratulatory page and scrolling down showed me the five photos I dated, gave averages for color and black-and-white as well as the actual dates, and then showed the five photos I hadn't been shown with the same information.
Interesting but for me on Chrome on Android, the results plots are completely off. My dot in 1992 was shown as later than the 2000 year mark, for example. So take care when reading the results.
I think there are two takeaways from this experiment. One is not at all noteworthy and one is somewhat interesting but has limited applications.
The uninteresting takeaway is that people note whether a photo is B&W/color and use that to help date the photo. Well, no kidding. We use a lot of contextual information to try to figure that out: not only B&W/color but also graininess, hair/clothing styles, architectural clues, etc. It should not be surprising that we do this. In fact, it would be surprising if we didn't do this.
The slightly more interesting takeaway is that we're bad at it, likely because most people don't have an accurate idea about when color photography became mainstream, and possibly also because some people fail to realize that sometimes B&W is an artistic choice, rather than a technical limitation.
As for what applications that might have ... well, I'm not sure. If it's extremely important (for, say, financial or political reasons) to accurately date a photo, my guess is that you'll bring an expert rather than a layman who knows to account for this. So really the only time I see the effect coming into play is when you have an armchair historian who doesn't know their own limitations.
I knew right off the bat they were messing with colorization in photos to try and prove that as a cause of date interpretation. However, I tried really hard to guess the context of the photos and map those to dates rather than mess with colors and clothing, which can be poor indicators given how cyclical style is. Biggest takeaway: it's really hard to accurately guess when a photo was taken unless you know what historical event it captured or there's a dead giveaway (like a piece of tech) in the photo.
particularly in the photo with the two "hippy" women - the only clue is how they dress, and that could be anywhere between 1965 and now, because retro fashion is a thing.
I think the experiment would be better if it accepted a range instead of a single date.
I saw that photo in color and missed it by +2 years. In color, at least, it's very obviously 'retro' and the phrase on the sign is not particularly 60's. I got two others within a year or two, but it was because I recognized the events. One I really flubbed was the two little boys on the junk pile (mine was color), which I guessed was related to the Fall of the Berlin Wall. Since the color average was '87 and the black-and-white '85, I'm guessing that's what most people thought regardless of color. The most difficult one for me was the 'street parade', which could have been many, many moments. In all cases, regardless of color or black-and-white I wasn't using whether the photo was color or black-and-white to determine the date. I was first trying to guess the possible events, and then trying to rule them out based on clothing, style of cars, architecture, etc. etc.
I think a bigger problem is where to initially put the selection knob. The knob on the picture of the two 'hippy' women was placed around '64 (if I remember correctly), confirming the bias in my brain "60s == hippy". I suspect most people link the selection knob mentally as a sort of 'caption' of the photo, even though they know consciously it's the selection tool.People are also probably slightly less likely to move a slider further than closer.
I was within two years of the right date on the hippy photo because of the design of the museum direction signs behind them. That clear acrylic and minimal font stuck to a wall style of signage is quite recent.
I am not sure if it is an accident, but there is clearly something wrong with the display/visualization of their results.
In 2 out of the 5 cases; I picked the year far more accurate than the avg., but the visualization of the dots on the timeline implied that the avg. was far more accurate than my guess.
For example; when my guess was off by one year and the avg was off by 4 and 6 years, the dot for the avg was right next to the actual date line, while the dot representing my guess was almost twice the diameter of the dot away from the actual date line.
It gives quite the false impression and is also quite an interesting example of distortion of reality and people's perceptions, even if not intentional. I theorize that most people do not look detailed enough or do the simple math to confirm that the visual representation is accurate; thereby leaving them an impression of accuracy and inaccuracy, depending on the results.
The results showed that colour had a large impact on people's guesses for some pictures, but for others there was very little difference. For me, colour didn't actually impact the guess much since I don't know when colour photography became mainstream, and even then, you get B&W pictures even today.
The hardest part for me was having no context of where the pictures were taken (although I should have guessed in the US). The picture of the farmers & the horse could have been taken today in many parts of Africa.
I'm not sure what this was supposed to demonstrate. It seems wholly without value. Period.
Here are some pictures of people, behaviors, clothing and facilities that likely existed before you were born. Some of these things probably still exist in various locales.
22 comments
[ 2.6 ms ] story [ 61.0 ms ] threadWhat surprised me is the magnitude of the effect in photos that were clearly taken after the advent of widespread mainstream colour photography. I'd have guessed the difference between the colour and non-colour groups would shrink considerably since colour would be much more a reflection of artistic choice than technological or resource constraint.
I learned I don't really know much about clothing styles.
From the site:
>When you were dating the photos, you probably looked for context clues — what people were wearing, if there were any familiar buildings, and if you recognized any faces. You were probably also looking at color.
I would argue that the very nature of the site makes it more likely that most people were using context clues. I also wonder if color gives you more detail/clues and hence in most of the photos I was shown, the color guesses were closer to the actual year.
I would find the accuracy of the guesses (as opposed to the difference between color and B&W guesses) more informational, for a larger data set.
It's actually an interesting point that is directly related to the subject research. How can you get an accurate picture when things cannot be discussed freely … you simply can't.
I encourage everyone to diversify your channels for for discussion and knowledge, be it distributed forums or even things like using some of the browsers and/or plugins that allow for open discussion on top of what is permitted by the heavy handed types (to put it it extremely lightly) among us.
The Einstein photo, for example... I knew he died in the 1950s, but I wasn't sure when.... but the audience was mostly Black, so that pushed it forward for me into the 1960s... without those clues, I would have been more wildly guessing.
I tried really hard to ignore the monochrome/color aspect... it would be really nice to know how far off I was on each after the fact.
The uninteresting takeaway is that people note whether a photo is B&W/color and use that to help date the photo. Well, no kidding. We use a lot of contextual information to try to figure that out: not only B&W/color but also graininess, hair/clothing styles, architectural clues, etc. It should not be surprising that we do this. In fact, it would be surprising if we didn't do this.
The slightly more interesting takeaway is that we're bad at it, likely because most people don't have an accurate idea about when color photography became mainstream, and possibly also because some people fail to realize that sometimes B&W is an artistic choice, rather than a technical limitation.
As for what applications that might have ... well, I'm not sure. If it's extremely important (for, say, financial or political reasons) to accurately date a photo, my guess is that you'll bring an expert rather than a layman who knows to account for this. So really the only time I see the effect coming into play is when you have an armchair historian who doesn't know their own limitations.
I think the experiment would be better if it accepted a range instead of a single date.
I think a bigger problem is where to initially put the selection knob. The knob on the picture of the two 'hippy' women was placed around '64 (if I remember correctly), confirming the bias in my brain "60s == hippy". I suspect most people link the selection knob mentally as a sort of 'caption' of the photo, even though they know consciously it's the selection tool.People are also probably slightly less likely to move a slider further than closer.
In 2 out of the 5 cases; I picked the year far more accurate than the avg., but the visualization of the dots on the timeline implied that the avg. was far more accurate than my guess.
For example; when my guess was off by one year and the avg was off by 4 and 6 years, the dot for the avg was right next to the actual date line, while the dot representing my guess was almost twice the diameter of the dot away from the actual date line.
It gives quite the false impression and is also quite an interesting example of distortion of reality and people's perceptions, even if not intentional. I theorize that most people do not look detailed enough or do the simple math to confirm that the visual representation is accurate; thereby leaving them an impression of accuracy and inaccuracy, depending on the results.
The hardest part for me was having no context of where the pictures were taken (although I should have guessed in the US). The picture of the farmers & the horse could have been taken today in many parts of Africa.
Even in a place that still farms like that with wagons that have wooden wheels, I would expect a baseball style cap instead.
Here are some pictures of people, behaviors, clothing and facilities that likely existed before you were born. Some of these things probably still exist in various locales.