Hacker News new | past | comments | ask | show | jobs | submit login
What Was the Happiest Day on the Internet This Decade? (theringer.com)
74 points by danso 15 days ago | hide | past | web | favorite | 49 comments

> There’s a lingering hope among people who are Extremely Online that the internet will somehow right itself and again become the place of bizarre daily delights that made it so engaging in the first place. Those days may be gone forever.

Meh. Just because a big part of the internet really is like the "torture chamber" of Clockwork Orange doesn't mean the internet of the early 2000s is gone. I used to hang out in some forums as a young teenager, and these forums are still there, and active. All of the IRC channels I was active in 2002 are still there. I can still use ICQ to send messages to people I now know for over 20 years. You can still spin up a personal homepage with nearly the exact amount of work as in 2001 (at least in Germany, you had to be very careful regarding legal issues even back then). It may be buried under a ton of shit (read: ton of tweets), but the internet of 2000 is still very much there.

Everyone likes to say you can go back and do things like they did in 2000 but it's never going to take away the fact that there is more stuff out there. Back in the day this "old internet" was all there was. There was no outside influence and the rules were all just being made. Now even if you go to old forums you can't help but see the influence they've taken on from the "new internet".

It's like growing up. Even if you're immortal you're only growing up once.

> a big part of the internet really is like the "torture chamber" of Clockwork Orange

Is it? My daily rounds almost never encounter anything like that.

Then you have either developed an immunity, or you are proofing my point: since it is possibly to move around in a sane, creative and/or interesting subset of the internet, the internet from 20 years ago is still out there :)

On a side note: this "good" internet may be hard to find, but don't forget that for the majority of the population even in highly developed countries, it was still uncommon to have access to the internet at home 20 years ago, so it was "hard" to find the good internet also back then. This is somewhat thought provoking, as it seems to indicate that the "good" parts of the internet were, and are, elitist circles.

I would have said “the day Pokémon Go had been released”.

I remember people from different generations walking in the streets, talking and laughing together without knowing each other. Something very rare for Paris :).

Is that really a day "on the Internet" though? That was beautiful precisely because everyone got outside, saw monuments and interesting sites, learned about their surroundings.

It was also entirely internet enabled. Worlds meeting, I suppose.

It was less fun in my town: pedestrians wandering obliviously into traffic, etc. Still entertaining though.

I'm coming in just to say this: Twitter is not the internet. There must be more complete places to pull this data, even if that takes more effort.

Yup, this data is the equivalent of dumpster diving to see what people are eating.

From the intro:

> let’s take a pseudoscientific journey

I have something better to say: the op is bs and don't make any sense.

Hilariously, the graph has peaks on new years. Presumably "Happy New Years" tweets are deemed quite happy.

Well you do need to be at least slightly happy to send a Happy New Years message, unless you're sending it in a sortof ironic depressive funk.

I think it's honestly more of just a meaningless statement (at least in the US). Sort of like when someone greets you with 'how are you?' and everyone just says 'great' or something like that.

It's just words to most people, I think.

But it's words they say because that's how they want to present themselves to total strangers; as a courteous person. In the same way, people could dress like total slobs, but instead take time every morning to look decent.

Okay, so we now defined the meaning of "being happy" to mean "not being so totally depressed to not even put on clothes anymore" :)

Quite a low bar to reach, we can do it!

Azure Text Analytics says "Happy New Year" has a sentiment of 96% - whatever that means.

Interesting that many of the big peaks don't have any specific meaning attached to them but the big troughs do.

I interpret this as meaning the internet needs a good reason to be sad, but can be happy for no particularly big reason. This seems like a good thing.

Likely something happened that is politically incorrect to even name, the bias is incredible given the troughs for tsunamis and nuclear plant meltdowns are a mere blip compared to Hillary not getting elected.

Also I'm not sure how/if they filter out bot traffic. Twitter, as I understand it, is mostly bots and fake follower accounts with human traffic being a tiny fraction of the data. Kinda like the old days of Usenet where 99% of the traffic was computers sending binaries to each other and only a tiny fraction was human to human text discussion. I would imagine the happiness sentiment of a mere repetitive shell script could distort the data such that the average is highly compressed around the 5 result. Possibly if we assume a nuclear plant meltdown should have scored a 1 on any realistic and reasonable human scale, then dropping only a tenth of that might indicate twitter is, to one sig fig, maybe ninety percent inhuman repetitive bot traffic.

To some extent twitter is workforce automation for echo chambers and signalling group affinity and membership; I'm not sure the mood of an environment in that context means anything when related to the real world, which reduces the significance of horrible events resulting in small signals. Surely, tens of thousands of deaths and nuclear contamination have nothing to do with trying to fit in with the cool kids, or at least a tiny subset of the population who think they're cool kids, which is a counterexample to my mathematical model in the paragraph above.

Also noticed this, FWIW I believe some of the recurring peaks are named once and label omitted in subsequent years. I.e, the higher peaks before every year are Christmas/New Years with Valentines Day shortly after.

The news article is misrepresenting several things:

1) Obviously, Twitter is not representative of the whole internet and its users. Twitter has a particularly activistic base, so it's no wonder "happiness" went down after Trump got elected.

2) It misrepresents the performance of the sentiment analysis method used:

This is based on sentiment detection research done in 2011 explained here: https://journals.plos.org/plosone/article?id=10.1371/journal...

They used a fairly simple lexicon-based sentiment analysis approach which even in 2011 was not even close to the state-of-the-art. They first create a sentiment/subjectivity dictionary by letting MTurkers rate words in different categories such as "happiness", "laughter", "greed", "hate", etc. For the final "positivity" score they add the matching words in Twitter posts.

This is a simple and efficient approach but produces fairly unreliable results w.r.t. negation, valence shifting, etc. Even in 2011, machine learning approaches were much better usually.

Regarding twitter, the population growed and changed composition over time too. For example pre and Post Oprah around 2009 and explosive growth until 2014 when it slowed somewhat. So even if it's not a good representative of the whole internet, it's not a good representative of itself when looked at over time. One would have to use the same cohort of users from each year with the assumption that their usage stayed the same.

And of course it's limited to english, so far from "the internet".

And indeed, Trump has been elected for getting the majority of votes, yet it represents it as a sad day, while I'm pretty sure most voters were happy.

Hillary won the popular vote. Also non-voters (including younger people who are ineligible or just didn't vote) and people outside of the US have opinions on such things.

I feel that Clinton winning the popular vote is also a statement that does not give a clear picture, iirc, her popular vote majority was entirely due to her humongous majority in the single state of California, if you exclude the outlier of that single state and consider the figure from the remaining 51 states, then Clinton no longer has a majority on the votes.

California is over 1/10th the US population. Why would you just ignore it?

Their electors got to vote. Same as how they have two Senators like every other state.

And let's toss this question back at you: why does California routinely ignore all of California outside LA and San Francisco?

Similarly, why did all the polls and candidates (including Republicans!) other than Trump entirely ignore the rust belt? That's why his win came out of nowhere, after all.

Obviously those areas are less populated, but why should a minority not have representation? Larger picture, how do you manage a union of states when not even states but a handful of cities hold all the political influence?

If we switch to a popular vote, a handful of coastal cities will have a lock on the federal government.

In the long run, it's called the United States because states are sovereign political entities, and that union works because small states have political influence. Instead of eliminating the electoral college, Democrats would do far better if they got back to their roots as the party of the working class.

A handful of cities don’t hold [all] the political influence. For things like what was being discussed, the presidency, cities don’t hold any more power than the rest of the country. People’s votes in NYC for example are pretty useless when voting for President.

A lot of your other things are pretty random and presumptive. You didn’t even agree that California should obviously count when considering who gets more votes nation-wide.

You're making a lot of presumptive, imprecise, and careless claims. I disagree with your conclusions, and I question the validity of your assumptions.

If you want to have this discussion, I'd urge you to engage in a more digestible and exacting approach, rather than trying to ramble your way through the topic.

This thread just seems like changing the goalposts. It started with someone saying that Trump won the majority of the votes. Which is just not true.

That is all

It also doesnt include the, relevant, fact that half of the USA did not vote. People seem to hear this as more Americans voted for Hillary than didnt. This isnt true. More Americans voted for Hillary than Trump. More Americans voted for no one than either candidate.

That's only true if you include Americans ineligible to vote (i.e. children). 61.4% of adult U.S. citizens cast ballots in 2016. More Americans would could vote for either candidate did so than not.

I doubt there's ever been a candidate that has won votes that total more than 50% of the voting age population so there's no reason to single out the results of the 2016 election.


A majority of adults have voted in every single American election if I’m not mistaken. This is including fringe people who might not have easy accessibility to voting too.

Um, I'm in the UK, but I thought Trump didn't get a majority, and in fact got less votes than Clinton; it was the electoral college system that put him in.

Wikipedia confirms that, https://en.m.wikipedia.org/wiki/2016_United_States_president..., 46% Republican, 48% Democrat.


Reality distortion on the right. Trump has falsely claimed (lied) that he won the popular vote, so his supporters parrot it, when it's clearly not true.

No. 1 'Happy Days', June 25th 2015

Obama legalizes gay marriage in the US,

but also,

Five different terrorist attacks in France, Tunisia, Somalia, Kuwait, and Syria occurred on what was dubbed Bloody Friday by international media. Upwards of 750 people were either killed or injured in these uncoordinated attacks.

You're conflating two different things, the U.S. Supreme Court legalized gay marriage (a happy event for many, sad for others), Obama sang "Amazing Grace" at a sad event, a funeral.

That's kinda negligible compared to the other 150k daily deaths around the world. It sure registers as a big event in the media, just like plane crashes. Probably because of saliency bias.

Accidents and deliberate acts are not comparable in my opinion. Accidents can be caused by negligence or just bad luck. On top of all this, humans killing other humans is just barbaric.

As determined by some weird algorithms based on Twitter data.

The happiest day on the internet is whichever day had the least amount of traffic.

I would have guessed Obama's first election win night. This was massive.

Seems a bit curious that "death of Osama bin Laden" shows such a strong dip. Much sadder than "death of Robin Williams", apparently.

One of them chose to die by their own will. The other was unlawfully assassinated in a perverted lust for revenge. I would be glad if the death of the latter was made society unhappier when comparing the two. Not so say that this is a morbid, perverse thing to do anyways.

This article is kind of trash. It basically starts off by saying you can write off anything that happened after 2016 because Trump winning made twitter go into a downspiral of sadness, when in reality that downtrend seemed to have started in mid 2015 (right around when election coverage started really ramping up, interestingly enough).

Then after going through the effort of showing graphs and numbers to make the rankings seem legit, the author describes how the scoring is done, and it seems incredibly skewed by the addition of arbitrary events that each add a point to the score. So basically, if the author wants a day to get a higher score, they can just look into it more to find more events so that its score goes up. If an event happened that day that contributed to happiness, the hedonometer score already reflects that, it makes no sense to double dip by giving the score extra points since that days happiness may have had multiple small contributions rather than just one big one.

I don't even really agree with how the hedonometer score was integrated. I think to really see how "happy" an individual day is, it would be better to subtract that day's happiness from the baseline/average happiness of that month rather than the raw score. This gives small happy days in a happy period an advantage over big happy days in a less happy period. I guess if you're just trying to say that day is happy without trying to tie any event to it, the raw data is fine, but I think that's much less interesting. There is probably some fancier math that could be done to weight the scores on a curve that would be more fair than just subtracting from their baseline if you really wanted to get into it.

In the end they come to the conclusion that same-sex marriage being legalized is the happiest day on the internet, but if you look at the first graph, it really barely looks like an out of the ordinary day, you wouldn't notice it if it wasn't being pointed out. This really makes it feel like the author came into this article wanting that to be number one and made everything else fit around that.

reading the tweets while Trump started to win the elections.. before was like "no way it will happens" and after just "omg he is winning", was kind of a tragic comic day.

November 8th 2016

Seriously. I was giggling like a little school girl all night long.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact