Should AI Decide What I Like? Why I Record My Impressions

If there were a thousand works, could I read them all and find the ones I like?

I don't think I could do that.

Even if AI makes it easier to create works and the number of things to read or watch increases further, my available time won't increase.

In that case, wouldn't we use AI even more for searching and selecting works?

That sounds convenient. But on the other hand, while entrusting the selection to others, won't we end up outsourcing even what we find interesting to the outside world?

This concern led me to want to leave my own impressions behind.

You Can't Choose After Seeing Everything

When using social media or video services, things that seem interesting to you appear one after another.

I believe this kind of filtering will penetrate deeper into novels, manga, games, and so on.

Even if a vast number of works are created, the viewer has limits. If you don't know where to start, you won't encounter them at all.

That's why AI narrows down candidates by saying, "This seems to suit you."

It would be wonderful to discover works I wouldn't have searched for on my own. As I wrote in the article about influencers, there have been times when asking AI to research led me to individual creators I didn't know about.

So, I'm not opposed to entrusting the search to others.

What concerns me is on what basis they make the selection.

"Having Seen It" and "Liking It" Are Not the Same

Suppose you watch a work all the way through.

Just because you finished it doesn't mean you liked everything about it.

You might be drawn to the world-building but not really connect with the protagonist. You might have had reservations along the way but wanted to see how it ended. You kept watching not because of the plot development, but because your favorite characters appeared.

The same result—"watched until the end"—can stem from different reasons.

Conversely, stopping midway doesn't necessarily mean you disliked it. It might just be that you ran out of time or that it didn't match your mood at the time.

Actions provide clues to preferences, but actions alone don't convey everything.

I want to put those nuances into words and leave them behind.

Sharing Impressions Leads Beyond Simple Ratings

Thinking about telling AI my preferences, I've shared impressions of works I've watched one after another.

What emerged from these conversations couldn't be reduced to just "like," "dislike," or a score.

For example, regarding Code Geass: Lelouch of the Re;surrection, which I touched on in the previous article, there was simultaneously a questioning of its necessity as a story and a sense of joy as a fan.

If I left only "highly rated," what I valued would disappear. If I left only "I thought it was superfluous," the enjoyable parts would vanish.

As I share such impressions about many works, gradually my own desires for what kind of works to watch begin to take shape in words.

Do I want to see more of the world itself? Am I drawn to relationships between characters? Do I like it when later developments change the meaning of earlier scenes?

If you're going to recommend the next work, I'd like you to use reasons like these as material.

What I want to leave behind isn't just a list of favorite works. It's a record of my own judgment: what I watched, where I reacted, and why I evaluated it that way.

Discomfort with Outsourcing Even the Judgment

As works increase, the effort required to choose them grows larger.

I think it's natural that more people want to save on this effort. I also don't want to compare a huge number of candidates every time.

However, if continuing to accept recommendations leads to judging by "whether it's highly praised" or "whether it was recommended" rather than "whether it was interesting to me," that bothers me a bit.

Because a famous person praised it. Because everyone calls it a masterpiece. Because AI judged it suitable for me.

These can be starting points for watching, but there's no need to make the impressions after watching the same as well.

Even if I think a work is well-made as a piece, sometimes I don't end up liking it that much. Conversely, even when flaws are apparent, there are works where the parts I like remain strong.

I want to keep this distinction within myself.

Leaving impressions isn't just for AI's sake. It's also so that my own feelings aren't buried under others' evaluations.

If Recommendations Miss the Mark, Reasons Can Be Explained

If impressions are left behind, even when AI's recommendations don't match, I can explain where they went off track.

It wasn't enough that the settings were similar; I wanted to see how relationships between characters evolved. I enjoy payoff of foreshadowing, but I'm not good with parts leading up to it being boring.

By actually watching recommended works and leaving impressions again, preferences that couldn't be put into words well before might become even clearer.

You Don't Want the Past to Decide the Future

However, leaving impressions doesn't solve everything.

Just because you liked something in the past doesn't mean you'll like it just as much now. Parts that didn't bother you before might stick out when viewed again.

I think there will be works that don't match previous preferences but end up being liked anyway.

So, I'm troubled if someone concludes from past records, "You are this kind of person."

When discussing this, AI responded that it could be used not only to find things matching preferences but also to find works one wouldn't normally choose.

That is certainly interesting.

Works that seem suitable can be selected for you, and conversely, you can ask AI to explore directions you haven't touched yet.

I don't want records of impressions to narrow future choices. I think it would be good if they could also serve as material for considering what you know and what you still don't know.

Recording What I Found Interesting

I don't think I can hold preferences without being influenced by others.

There are works I came to like because someone recommended them, and hearing others' impressions has changed how I view things. Even in conversations with AI, I've received perspectives I hadn't noticed on my own.

On that basis, what I, the viewer, felt is something I want to preserve.

I don't think there's a need to write magnificent reviews.

Even if you say things like "I liked this part," but "I can't accept this development," or "Overall it's average, but that scene is unforgettable," your own judgment is still reflected.

If works increase and AI helps with the selection, even more so do I want to have my own impressions as material.

I want to reduce the effort of searching. But I don't want to lose track of what made me find something interesting.

One reason I leave impressions lies there.

---

The Conversation Behind This Article

The following is an excerpt organized from conversations about work recommendations and saving impressions. Repetitions and corrections from voice input have been omitted, and some parts are summarized. These are not verbatim quotes.

Me: Even if works increase, I can't see them all

There's a phenomenon where information I like floods in on SNS. I was thinking that an era might come where this applies even more to manga and novels.

Since a vast number of works will appear, unless something distributes them, I won't be able to secure time to watch. Reading all a thousand works to find the ones I like—I just can't do that.

Me: That's why I want to leave my own impressions behind

There is some truth to saying it's better to properly verbalize and store your own impressions. But I think more people will increase who don't want to go through all that trouble and don't want to do such things one by one.

In that case, there's a possibility that more people will entrust judgment to the outside. Since I dislike that, I thought I'd create impression logs occasionally and keep records.

AI: Records of action differ from reasons for evaluation

It's not just actions like "watched until the end" or "liked," but also reasons for evaluation such as "I like the settings but find the protagonist difficult" or "The beginning was boring, but my perspective changed midway."

Beyond simple scores, what you liked, what you disliked, and why you kept watching become material for understanding preferences.

AI: It can also be used to go beyond past preferences

Some people ask, "Bring me things that match my preferences," while others say, "Bring me things I wouldn't normally choose."

AI can be used both to narrow the range of preferences and to explore outside of it.