August LSAT Parallel Reasoning rolled out lots of "multiple-flaw" stimuluses with answer choices that never quite matched, including some flaws and excluding others.

It was up to us to determine which flaws were "the biggest sins" and worth accepting compared to the other answers, in order to choose the best answer.

Is there any guidance on how to "weigh" these flaws against each other?

Currently, I think it's probably something like:

Conclusion Mismatch > Formal Logic (including Quantifiers/If) Mismatch > Modality > Fallacies > Terminology Mismatch

But this is just how I would do it and I have no idea if there's a better way to do this. A lot of parallel reasoning has tried to make itself diagram-proof on the test by requiring you to weigh flawed answer choices against each other, and I was hoping for some guidance on this.

Other things i noticed from August was that there are some supercharged PAI questions.

It is using a crazy amount of tokens and being a lot more analytical, I've had it for a few days and have already used almost all of it with way less queries than before.

Bit by bit, I feel like I am decreasing my variance.

I made a post on here after June's exam, talking about how a lot of the exam took me by surprise. I put my observations in the post, but it was mostly about how the entire thing was a diabolical phenomenon-hypothesis + quantifier + causal reasoning + correlation bomb.

I have worked on strengthening each area that gave me pause, and I think I am better than I was in June. But it is impossible to tell. I scored a 178 today, a 180 on my test before that. Only one question on the Ptest I took today reminded me of what I saw in June (it was a phenomenon-hypothesis question about waters surrounding Shooter's island, which thankfully I have grown wise enough to get right).

But technically, every test I take now is either some % not fresh or I haven't seen the content in a very long time. Literally no score is "representative" and I feel like I could get absolutely decimated on the exam, even though in every "fresh" metric I am mogging my June self.

I got a 168 in June despite multiple mid 175 performances. What has changed between then and now is the confidence I go at questions with. I assume the way I interpret words, negatives, phrases, etc. is always correct and I don't look for "safer" answers. I hope it pays off. But I may very well be shocked again and join y'all this October.

If any tutors/instructors see this, do you have any thoughts about how I should interpret my testing data?

Intro

I have been using the AI tutor since it came out to study for the August LSAT exam, and here are some of my main thoughts on it.

Usage Limits

I wish there was a warning or something when you're close to reaching the AI tutor usage limit. I never used it for images or anything extremely intensive, but I think a monthly limit as opposed to a daily one like other AI tools is a lot more unpredictable. Given that the subscription is $69/month at a minimum and the tutor isn't really useful for reasoning about the actual questions, I think the usage limit should either be more generous or reset more regularly.

If you still want a way to sell it as a feature for higher tiers, I think stuff like image access would be a better selling point, since people who are paying for that tier likely are more willing to pay for tutoring anyway.

Good Features

1) It is helpful that it can do things like make drills featuring questions you have not seen in X amount of days, but it does make me wonder why exactly users of the normal drill builder don't already have that feature.

2) It has a good memory. I have given it a list of mistakes I tend to make and asked it to categorize my mistakes based on my WAJ write ups for each given question, and it does a pretty good job of categorizing what exactly went wrong. This can also be a problem, because sometimes it feels obsessed with trying to show you that it remembered what you said, like it read a book on how to get people to like it or something.

3) It's a good motivator

Major Flaws

1) It has no concept of time, but it keeps trying to assure me that it knows what time and day it is.

2) It can be quite the sycophant, but I appreciate that there is a "coach" voice where sometimes it senses you are making a compromise or doing a drill that is below your skill level. Who else would notice I have done 4 sessions in a row without touching RC?

3) I have no idea how to tell why it believes anything it does. A personal issue I was having was that I was not diagramming difficult parallel flaw questions because I was following the wisdom that one should not diagram unless it is absolutely necessary. The AI tutor said something like the following: "Oh yeah, classic 170+ scorer thing, they burn tons of time refusing to diagram because of their pride and actually use up more time than if they had just diagrammed."

Do people ever do that? Maybe? It sounds plausible? But like, I have no idea if that's true.

Hey folks,

I got my June score back, and it was as I expected based on the difficulty of the exam. However, I’ve taken nearly all of the “modern” LSAT practice exams and all of the recent ones. I am kind of puzzled as to how to prepare for the real thing, because it has become very different from the last exam administration we have materials from (2020).

I was shocked by how time-consuming the questions were and the total focus on causality/phenomenon-hypothesis over anything else. Nothing I’ve ever practiced with feels remotely comparable, so I feel like all the drilling I do is on easy mode.

June exams were all new content (no reused content other than a single RC passage) which could explain the new style, but I haven’t really seen anyone else with this take, so I might just be hallucinating.

Does anyone have any advice on how to prepare for August?

My observations:

1) Parallel reasoning without formal logic to make the answer choices less strong

2) High amounts of wordy causal reasoning and phenomenon hypothesis questions with much more intense answer choices than I’ve seen before

3) A much harder time crunch on LR, but RC has remained the same

4) An emphasis on quantifiers/math style questions

My understanding is that drills have a "share" system intended to allow people to create drills for particularly hard/helpful content, helping people build specific skills outside the built-in difficulty/tags system we can use on our own.

However, I don't think I have ever seen anyone other than 7Sage tutors use the drill link feature, so I was thinking maybe 7Sage could start making their own sets. Maybe they could even be used to "introduce" people to making their own drills.

For example, when preparing for June, I selected all the passages from the PowerScore crystal ball and manually organized them into 4 separate drills based on PowerScore's categories. This was somewhat tedious, and I bet other people did something similar.

In case I have to take August, I'm currently trying to make a drill that focuses on math problems, but I find that the math tag + raising the difficulty isn't really representative of the math questions most folks think are hard. They're tricky questions for sure, but not because of the math.