Post by @punkess • Hey
Thank you so much for all your comments and discussion around the "`lens score`" .
**PLEASE SHARE YOUR THOUGHTS** on 3️⃣ 👇
1️⃣ I'll reply to everyone who
Comments
- Hi Alice! Can we check mine?
- Interesting approach, thanks for sharing! I question how effective this type of scoring is when it seems fairly subjectively based on collecting and publishing collectible posts.
Anecdotally, I imagine my score would be quite low since I've never sold a collectible, and haven't collected many. I also have nearly a 10:1 comment-to-post ratio, since I rarely have things I believe are worth posting about, but often feel that I have something to contribute to discussions (like this one!). My personal belief is that I'm creating noise/spam if I post, just to post. I don't think anyone would suspect me of being a bot or spammy. As such, it seems like this scoring perhaps suffers from confirmation bias, and may not be be useful for more than vanity metrics.
Identifying bots and spam is difficult, and getting harder every day. It's something that has been the topic of substantial academic research and even the largest social networks are unable to effectively keep it at bay. A good Lens Score would be one that attempts to recognize that people use social apps in many different ways, and allows for more than one type of user to have a "high" score. It's important to note that Lens can be used in many ways, ways that don't resemble a typical "social network", so it may be prudent to only include stats from known App IDs (like Hey, Phaver, Orb, Buttrfly, etc).
I also believe it may be more effective if a 1.0 represents suspected bot/spam, not the other way around. Proving personhood is basically impossible without some sort of biometric proof, while proving likelihood of a bot is difficult, but generally achievable (at least for now). I think it also comes off as less of a vanity metric when the perceived goal is bot identification, not ranking how *good* of a user you are.
The biggest issue, however, is that the moment you publish your formula it's no longer useful. Bots and bad actors can quickly, and easily, adjust to fool your system. This means we ultimately have to trust your judgement, or only use these types of scores retrospectively.
PS - This post and discussion is what I love most about Web3. ❤️
- What about me Alice?
- Gm Alice, that's really cool!
- Nice work! Pretty simple scoring rules that divide up the account space into a reasonable distribution 👍 👍 👍
- Hi, gm, Alice ! What's my score ?
- Very cool, I know this is super early, so there doesn’t have to be an answer. But how does this scale (or does it) as the people generally have more paid collects/collectors?
- nice, maybe adding weight for mirroror whose paid mirrored post gets collected to incentivize paid mirrors that are successful
- The way you calculate the score is very creative👌👌
What's mine ?😁
- Great start. I'll look into it a bit later to see what can be improved.
So, what's my score?
- Hi Alice! Could you please share my score?
- GM @punkess.lens ☀️. What's my score? Hoping it's north of zero😶
- Simple & efficient scoring system!
Personally, I haven't started doing paid collects which is why my score is just shy of 0.5 (it's 0.47). Impressive still that a 3 month old account can have such a high score only based on organic engagement. So, the criteria is definitely not skewed towards those who've been here longer!
Looking forward to this being built out!
- This is a decent start but coming up with heuristic scores has its own limitations. May be a simple learning algorithm with the signals you mentioned would do a better job :)
- Can you check me too pls?
- Dear Alice can you tell my score please?
- Amazing initiative, Alice!
This is one of my favorite use cases for Machine Learning and data processing, and I'm currently working on a more complex solution that I would love to share with you as soon as I'm back from my vacations in a week (I'm in Chile right now). However, I can already add some insights to your study.
📝 **Content Classification:**
To score a profile, we should first classify its content. The simplest variables I would consider for that are:
👉 **Keyword Analysis:** This involves identifying keywords such as "claim," "earn," "airdrop," "don't miss," etc.
👉 **Recommendations:** Good content can be classified based on how many (good) profiles have shared it with others. In social media, we can adapt the engagement rate (engagement/followers) for this purpose.
👉 **Spam Reports**
Paid and sponsored content should be excluded from the analysis above as they have their own dynamics and should be treated separately.
👩 **Profile Classification:**
This topic deserves a more in-depth discussion, especially considering transferable profiles and editable handles. I believe what we're really looking for is a profile manager's classification, tied to the wallet address behind it.
But I would love to hear others' opinions about it. I recall someone suggesting a score reboot after transfer, and this could be a good idea too.
Regarding the classification parameters, here are some suggestions:
👉 **Good Content / Bad Content Rate**
👉 **Mentions / Followers**
👉 **Average Engagement Rate**
👉 **Paid Collects Received / Followers** (Note: This parameter shouldn't heavily influence the score since not everyone is a paid content creator. It's a plus, not mandatory.)
👉 **Spam Reports**
These suggestions draw from traditional social media techniques and SEO practices.
- This is astounding work. Thank you fren.
- Super curious to see my score
- You've really put in a lot of work into this.
Well done 🙂
- @punkess.lens I can deploy an UI to getting your score based on this wrguments if you want 👀
- How do you find out your score?
- @punkess.lens maybe some poap for the og 1.0 people 🤣
- My initial thoughts:
Comment post ratio could be easily gamed by bots. All they have to do is keep the ratio between 1 and 10 to be regarded as real.
Number of paid posts collectors is determined by whether a person wants to enable paid collects on their posts, which doesn't really have anything to do with them being real or a bot.
Number of paid collects is affected by how much money a person is willing and able to spend on Lens, so it will put some people at a disadvantage through no fault of their own.
- 🤔 What do you think of this score alice?
- GM Alice , what's my socre 😃🤔
- What s my score please?
- oh that's look awesome.
- Can you check mine?
- How to see the score?
- Excuse me, how about me?
- Very cool, I know
- score is meaningless until we know the methodology
- What’s mine
- Good jobbb
- good analysis
- Still have to pay
- how to know my lens score?
- I want to see my score
please share 💚
- so paid more will get higher score?