Your Critical Thinking Problem Is a Culture Problem

Transcript

0:00

Junior: Your team thinks just as well as it’s allowed to disagree,

0:00

Junior: which might not be much. So they may not think so well. The trend that we’re seeing is that critical thinking is being outsourced a whole lot faster than it’s being replaced. What are we seeing in our organization with the organizations that come and partner with us? We’re seeing fluent answers that nobody can defend

0:00

Junior: because they weren’t theirs. The decisions with no visible—

0:00

Tim Clark: I have to laugh at that one, Junior. It’s true.

0:00

Tim Clark: A fluent answer that no one can defend. Where did it come from?

0:00

Junior: Yeah, you push on it for two seconds and the person has no recourse

0:01

Tim Clark: because they can’t explain the causal chain. They didn’t do the homework. Oh. They didn’t do the hard, hard work to come up with that. No, no, it’s very, very shallow.

0:01

Junior: You’ll see decisions made with no visible reasoning and teams that eventually quietly lose this ability to push back. And so you’ll see the fluent answer with no reasoning behind it that just flies through the decision-making process because no one’s gonna push on it because they haven’t done the reasoning. The person on the other side hasn’t done the reasoning. So we just see a frontier model spit out X, Y, or Z and we say, great, carry on.

0:01

Tim Clark: And then you may have a culture that allows you or does not allow you to push on it. That’s right. But in the first place, did you do the work? Did you do the critical thinking? Did you wrestle with this to come up with a point of view, put something on the table, whatever it is, to be able to defend it or to critically think about it?

0:02

Junior: Yeah. So the second order claim in today’s episode is that the thing that keeps a team’s critical thinking alive is not the individual intellect of the team members, but rather challenger safety. The ability for the team to feel safe, disagreeing, challenging the status quo, asking a pointed question, improving something, making a critique, that’s the idea.

0:02

Tim Clark: I wanna just qualify that a little bit, Junior. You have to have your own critical thinking motor. Yeah. But what you’re saying is that that motor, even that motor is being threatened in the current environment because of the tooling, because of what’s available in technology.

0:02

Junior: Yeah, and we’ll look at the literacy data later.

0:02

Tim Clark: But that in and of itself is very frightening because what we’re talking about is moving employees from actively thinking, critically thinking agents into passive objects that are really not doing the work.

0:03

Junior: Well, and to add fuel to the fire, the cost of not thinking well has increased. So the cost hasn’t stayed the same for critical thinking. The risk has gone up because the speed of execution has increased and the cost of execution has decreased. So you can screw up an organization really fast. So you make the wrong call, because the execution cost is low, you can then implement that call overnight and cause really big problems.

0:03

Tim Clark: But I think for many employees, ostensibly, if you look at things, they think the opposite, because they are now armed and dangerous with AI agents that can help them that are, and so they are willing to outsource their critical thinking. They think they’re more powerful than they were, but they’re not doing the thinking. This is where it becomes very dangerous. And they believe that it’s okay if they outsource their critical thinking. So they think that they have become more capable when in reality, they’re becoming less capable because they’re outsourcing. Well, and at the surface level, you’d think it’s a reasonable claim. If you can give however many agents it was, a hundred thousand agents, a hundred billion tokens, $6 million in 88 hours and solve Navier-Stokes. Is that what it was? Yeah, I mean, in round numbers.

0:04

Junior: Then you might think, well, the thing’s pretty smart. Yeah, it is at computation, right? Yeah, it is. But is it as intelligent and evaluative cognition? Can it objective set the way that you need to? Can it weigh moral risk the way that you need to? Can it take responsibility the way you need to? No. And so if you wanna outsource computational cognition, do it, automate as much as possible.

0:05

Tim Clark: Understand the limits.

0:05

Junior: But if you reduce your total general critical thinking because you think that AI is going to do all of that evaluative work

0:05

Tim Clark: for you, you’re in trouble. You’re in big trouble. Yeah, you’re being diminished even as you think you’re being empowered by the AI. That’s right.

0:05

Junior: So what is critical thinking? It’s the disciplined work of framing a problem correctly, which is a large piece of what we’re going to talk about today.

0:05

Junior: Weighing the evidence and testing your own reasoning before you go into act

0:05

Junior: mode, execution mode. And a really interesting thing here, nobody measures that in a population over time. We don’t have good critical thinking metrics.

0:06

Tim Clark: We have some proxy indicators that we can look to,

0:06

Junior: but we don’t measure that across a population. So anybody claiming critical thinking has declined, we’re inferring. It’s really difficult to go and do that in any sort of empirical way. But here’s what the proxies do say. The OECD survey of adult skills, US adult literacy fell 270 in 2017 to 261 in 2023. So what’s happening? Our population is becoming less literate. In an age where it’s increasingly easy to become literate.

0:06

Tim Clark: In the age of AI, that seems paradoxical.

0:06

Junior: Yeah, and the share at level one or below, reading level one or below rose 19% to 28%. There’s a nine point drop in literacy across that particular cohort that already wasn’t

0:06

Tim Clark: doing well, level one. Yeah.

0:07

Junior: Literacy declined in 11 of 27 countries. And the pattern runs this way. Among 55 to 65 year olds, 12 countries declined and four improved. So in older population

0:07

Tim Clark: literacy is declining as well.

0:07

Junior: And among 16 to 24 year olds, eight declined and three improved. So still general decline. US 12th grade reading, 2024 against 1992, 22 year period down 24 points at the 10th percentile, down eight at the median, none at the 90th, which is super interesting to me. So that top 10% is staying there. The bottom is falling out. So in order to get ahead, you have to stay where you are historically across that 20 year period, the top 10% has been able to say, you know what, this is actually a pretty important thing that I learned how to read and write. And we see a decrease across the rest of the population

0:08

Tim Clark: across that time period. That’s scary to me. It is scary.

0:08

Junior: That’s very scary to me. So this decline predates generative AI majorly.

0:08

Tim Clark: Yeah, it does. Right? Yeah.

0:08

Junior: So what do you think is gonna happen in generative AI? Do you think that the average person is more or less willing to outsource

0:08

Tim Clark: their literacy to a model? More. More, of course. It’s just gonna accelerate the decline for most people.

0:08

Junior: Well, and anecdotally in my conversations with people at higher education institutions, you look at the incidence of AI in anything literary.

0:08

Tim Clark: Yeah. Up, it’s gone up.

0:08

Junior: And real power users who are real power critical thinkers and power creators and power technical users of AI, I mean, top 1% of the 1%, they’re taking huge advantage of this because they’re automating the things that should be. [09:50:47 - 09:50:48] Tim: Should be automated.

0:09

Junior: In computation.

0:09

Tim Clark: Yeah.

0:09

Junior: And they’re keeping the things that should be uniquely there.

0:09

Tim Clark: So— So overall, it’s amplifying their overall capability and skills.

0:09

Junior: When we talk about this in leading through AI about the difference between human and machine and the things that we need to keep that are uniquely ours and those that we can outsource and delegate to the machine. So I would say that at the outset of the conversation, that’s one of the first things that you have to define for yourself. What do you think is uniquely human? And what do you think fits into the category of computation that can just be delegated? If we don’t make it that far, it becomes really difficult to say, how should we use this machine? How should we use this tooling? We have to understand the landscape first. Because if you assume that that thing is generally intelligent across evaluative things too, then you can get yourself into a hole. So we often say that it’s better to make a decision than no decision, that it’s better to move.

0:10

Tim Clark: We talk about this idea of permanent offense. That’s kind of been conventional wisdom for a long time. I mean, we say that, right? Well, and you look at all the accident reviews with CEOs, right, Junior? And they come back and they say, what’s my biggest regret? I wish I would have moved faster. I wish I would have taken on this initiative and why did I hesitate? Why did I wait? Why did I move so slowly? We hear this over and over again.

0:10

Junior: But here’s what’s interesting about the new environment. If the marginal cost of execution is falling and the execution still

0:10

Tim Clark: creates path dependence

0:10

Junior: and the path dependence includes compounding consequences,

0:10

Junior: then AI allows us to get to some of those compounding consequences sooner. [09:52:36 - 09:52:40] Junior: And so there’s less lag in the input and output, which means two things,

0:10

Tim Clark: that if we make good decisions,

0:11

Junior: we can compound those good decisions faster. If we make a poor decision, we can compound those poor decisions faster. And so there’s an amplification that’s happening to our decision-making power. If we’re good decision-makers, AI will amplify that output for the better. If we’re poor decision-makers, AI will amplify that decision-making for the worse.

0:11

Tim Clark: So we can get into trouble faster.

0:11

Junior: Exactly.

0:11

Tim Clark: Or we can accelerate to compounding positive consequences faster.

0:11

Junior: Exactly. I think about how far away we are at our organization from just monumentally damaging the organization, like hours.

0:11

Tim Clark: Don’t say that. Really, it’s true. But it’s true.

0:11

Junior: It’s true. The power that we wield with the tooling that we use every day is such that we are hours away from something going

0:11

Tim Clark: catastrophically wrong. We used to think about path dependence and the compounding of potentially unintended consequences, Jr., as taking a long time.

0:12

Junior: Well, and even if you didn’t think about the bad thing happening as a bomb going off in the organization and just a really catastrophic problem, you could just make a wrong strategic move and move in just directionally somewhere that’s not going to be profitable for whatever reason, or that increases the complexity of the organization in a way that was unintentional. There are ways that allow you to veer off track that seem innocuous. A really good example is, let’s say that the cost of execution on a marketing campaign has gone down. So we could come up with some new messaging and we could ship it tomorrow.

0:12

Tim Clark: Yeah, today.

0:12

Junior: Today, yeah. What would happen? Would that be catastrophic tomorrow? No, it wouldn’t. But would that confuse our buyer? Yeah, probably. How much effect would that have down the road? A lot. So if I sat down with a model, created a whole campaign, skills we have access to, the distribution we have access to, gave it budget and flooded our market with poor messaging, that becomes really problematic. It’s not catastrophic tomorrow, but that compounds really quickly because of the exposure I’m able to gain through AI. Let’s say I start doing messaging and things of that nature with some of the tools. I could really cause some big problems.

0:13

Tim Clark: Here’s another implication though, Junior. If you think about it, if you ask anyone, hey, we’re going to try, we’re going to come up with a new marketing campaign. We’re going to launch that. We’re going to try this, that. Most everyone would agree that that is a one way, or that’s a two way door. As a decision, okay, go try it. And if it doesn’t work, come back, we’ll adjust, we’ll make changes, that’s fine. I think what AI is doing, as you said, is the marginal cost of execution goes down. That means that we have acceleration through the execution. That means we get to consequences good or bad faster. So then what does all of this mean? It’s blurring the line between a two way door decision and a one way door decision. If you can create this campaign and go execute it quickly and create the consequences and the path dependence, well, if they’re negative, if you’re compounding negative, adverse, unintended consequences, you can get down the road very quickly, suddenly your two way door decision looks more like a one way door, how are you going to get back? So that’s something that AI is doing, is blurring the line between a two way door decision and a one way door.

0:14

Junior: Speaking of two way and one way, let’s talk about system one and system two thinking. We often say that system one is good for the two way door. And that system two is good for the one way door,

0:15

Tim Clark: we got to be more deliberate if we’re going to go through that thing.

0:15

Junior: So system one thinking is fast, it’s automatic, it’s effortless, it’s your first thought that arise just by itself. The system two is different, it’s slow, it’s deliberate, it’s effortful. We have a course in decision making using the decider model, which is about system two thinking. How do we become more deliberate as we approach more consequential decisions? Here’s what’s happened in this AI talk. What traditionally took system two thinking, now only requires system one, because we can outsource what should have been our own system two thinking to a model. And when it comes back with idea number one, we say, well, I guess something smarter than me thought about it, so let’s just go ahead and run with it. People who aren’t diligent, rarely get to that point of critiquing, double checking, thinking more deeply about that first answer that comes back from model. And so the diligence isn’t there, which hurts us on the framing end. So if we go in and we just take the first thing that comes back, we can get into a big problem. Execution was expensive historically, which meant that even if we made a system one poor decision, it wasn’t going to go very far.

0:16

Junior: That’s right. But now if we make a system one call on what should have been a more effortful decision, it becomes more problematic.

0:16

Tim Clark: So see how quickly we can become intellectually lazy. Yeah. If we’re outsourcing. Yeah. We are going to outsource the computational cognition. But as you say, Junior, there’s the danger of not discerning and discriminating between the computational and the evaluative and what we have to maintain a hold of, keep a hold of that, as you said, objection setting, assigning value, taking responsibility, some of those irreducibly human responsibilities.

0:17

Junior: Yeah. Well, two examples. One, I like the example of a website because the cost of execution is so low. If you just say, go spin up a site, the chances that you get a very beautiful wrong site are high. It’s going to look very good and it’s going to be completely wrong.

0:17

Tim Clark: Yeah.

0:17

Junior: And so that’s an area where we need system two, even though the cost of execution, maybe because the cost of execution is so low, we need to spend some time and say, what actually is this thing supposed to do? Take a training, right? People say, well, are the forces of commoditization eating at training? Well, a certain area, but again, you may get a very beautiful incorrect training because you haven’t had the type of disagreement, the type of framing, the type of pushback that you need. And as we’ve talked in this conversation so far, I think largely the frame has been me, a decision, an AI model, I’m doing the execution, I’m doing the evaluation, but we haven’t yet talked about collaboration and the fact that other people should be involved in this process is something that we need to do. And so that’s what that says, especially at the beginning.

0:18

Tim Clark: Especially at the beginning.

0:18

Junior: So we talked a few weeks ago about the fact that decision quality is not a function of time, it’s a function of information. How do we get the best information through collaboration and disagreement? And so this is where we’re going to turn the episode toward how do we get disagreement, right? One of the big critiques of models, really good frontier models, even 12 months ago, they’re sycophantic, they’re just agreeing with me

0:18

Tim Clark: all the time and I cannot get them to be critical for the life of me. Hey, that’s a good idea.

0:18

Junior: Even with prompting.

0:19

Tim Clark: Yeah, that’s right.

0:19

Junior: Today, it’s very much the same way. And so there is something that’s uniquely human still about the ability to sit down and deliberate with another person and have healthy intellectual friction and productive disagreement. So what we’re going to get to, the short of it is, that you will not make good decisions or the best decisions without a high level of disagreement.

0:19

Tim Clark: And you need to invite that into the process from the very beginning.

0:19

Junior: And you’re not going to get that disagreement unless you do that, unless you introduce it from the beginning, unless it’s aimed at, unless the environment allows for it, encourages it, you’re not going to get it. So why do we know that? Challenger Safety is the lowest scoring stage in all of our PS index data across 1.3 million data points.

0:19

Tim Clark: So that, it’s significantly different. It’s a 30 point difference

0:20

Junior: from stage one to stage four because it’s more vulnerable.

0:20

Tim Clark: So you’re falling off a cliff. And the reason that you’re falling off a cliff is because you have arrived at this place where you are now asking people to challenge the status quo, to engage in constructive debate, and the stakes are now higher, and nature of vulnerability has shifted. The activity that you are engaged in is a different kind of activity. And you need to make it safe for people to do that. You have to, you need to be able to create the conditions that will foster and motivate that kind of behavior. That is not an easy thing. And as we like to say, Junior, that environment is not found in nature. It is an unnatural state. Therefore, how do we do it? The leader is responsible to create those conditions. But there’s another complication, which is that the leader, is themselves a liability in that very process. And the leader must neutralize the liability of themself and their position in order to get to a place where the team is willing to engage in that kind of debate and dissent. And it all begins with, if we go back to the decider model, it all begins with the first step, which is define the problem, which goes to framing. Everything goes back to framing. If we misframe, everything that comes downstream is going to be off target.

0:21

Junior: Well, if someone were to take a contrary point of view and they might say, well, if a smart team or a high-performing team or a culture where people can be high performing is an unnatural state, that’s just because of the natural distribution. And I’m gonna get a range of intellect. And the unnatural state that I’m gonna put the team into is I’m going to cherry pick the smartest people, and I’m gonna put them in a room, and that will solve my problem.

0:22

Tim Clark: So it’s IQ points.

0:22

Junior: Exactly. And there are problems with that. There are many problems with that. One of them is that more often than not, you’ll get a smart individual with a private opinion. And what we need is a group of smart people with public opinions.

0:22

Junior: And the public opinion, turning an individual’s opinion

0:22

Junior: from private to public is the mechanism we’re talking about.

0:22

Tim Clark: That is exactly the mechanism.

0:22

Junior: And if it stays private, then you’re gonna have a whole group of individuals with private opinions. And the way that that problem gets framed, the way that that problem gets solved is going to be different than if we were sharing opinion. And that disagreement, which it will often end in dissent and disagreement, needs to be of an intellectual nature and not a social nature. And so it’s the leader’s responsibility to diffuse all of that upfront and create the conditions in which that can be true. That’s the unnatural state we’re talking about. We’re not saying, well, just beat the curve and go solve it with IQ. We’re saying that you could have a very smart, dumb team. And it’s not necessarily true that IQ and decision-making are tightly correlated. You’ve probably met some very smart people who make very poor decisions.

0:23

Junior: And it’s not necessarily true that if you just have a bunch of people that you’re gonna get a great result. There has to be this collaboration happening and stage four challenger safety that has to be there in order for those opinions to get shared [10:05:44 - 10:06:39] Tim: in a way that’s healthy. Junior reminds me of the finding that we’ve arrived at in our work with teams. And that is the team density does not equal, or talent density does not equal team performance. This goes right back to that point. You can put a bunch of geniuses together, but that doesn’t necessarily mean that they’re going to perform and that the quality of their output will be any good. It goes back to your premise. The team, your team will think as well as it is allowed to disagree. So now it becomes social in nature. It’s not purely intellectual. Now it has to be extruded through this social process where we are working together and we are helping each other through the process of intellectual friction.

0:24

Junior: Well, and you might go so far as to say that it’s not just a little bit social, but that it’s mostly social. And if you look at problem framing, which is where we’ll spend the majority of the time, that is ultimately important. So there’s some really interesting research from 99, from Nut, NUTT, 356 real decisions. They were two years each, half of them failed. They’d looked retroactively at the failure pattern. The most common failure pattern of the failed decisions was jumping to a ready-made idea at 37%. So 37% of the failed decisions could be attributed to just executing on the first thing that came across the table.

0:25

Junior: The first idea we have, let’s go for it.

0:25

Tim Clark: Yeah.

0:25

Junior: That is so instructive to me.

0:25

Tim Clark: If we know that

0:25

Junior: that’s the failure pattern, then we know that we can avoid 37% of failure by not doing the first thing we think of. How are we going to get to the second, third, fourth, fifth things we think of? We won’t if they’re private opinions. It must be true that they’d be public opinions if we’re going to get more than one.

0:26

Tim Clark: That’s right. So think about the tax on the organization that this represents.

0:26

Junior: So think about the causal mechanism in stage four. The causal mechanism is modeling and rewarding vulnerability. So what does that mean as a leader? It means that first, I have to model disagreement in a healthy way. I have to say, hey, I appreciate that idea very much. Let’s put that one on the board. Here’s another one. I haven’t thought this one all the way through, but here’s this one too. And then rewarding. Someone else puts an idea on the table. Hey, thank you so much for putting that idea on the table. Let’s consider that one as well. Great, we’ve got seven. Let’s proceed. If the manager doesn’t do that well and manage the social situation, highly unlikely that we’re going to get to a point where we have six on the

0:27

Tim Clark: table that are reasonable. Basically, I think what you’re saying, Junior, is that, and this is as you’re citing this work from Paul Nutt at Ohio State University, who’s a decision scientist, that teams think about if they can’t handle the intellectual friction in a social process, they short circuit the whole thing. They don’t even engage in the rigorous process of let’s frame the decision, let’s come up with criteria, decision criteria, let’s weight those, let’s think about alternatives, let’s see if we can evaluate those alternatives, let’s pick the best one, let’s test it, and then finally we’ll take it into the wild. They are short circuiting that entire process of the rational decision making model that we should be engaging in. Why? Because the team is unable and unwilling to go through this social process. It’s too painful, or they simply don’t have the capability to do it, so they don’t even do it. That’s why they grasp the first, as you said, ready-made decision that’s available, let’s go.

0:28

Junior: You may have someone in the room who’s been in a similar situation, who can forecast out the second, third, fourth order consequences,

0:28

Tim Clark: they know a decision’s bad and they still say nothing,

0:28

Junior: and they say, “I know where this is gonna go, “but we’re just gonna leave it be.” That’s so sad. We could avoid a whole bunch of pain and suffering if we just allowed, encouraged, built infrastructure for that opinion to be shared. So one practical thing that you can do as a leader is start to build the norm when you’re looking at decisions, when you are evaluating problems that are on the table, by saying, “How are we misframing this?” As just an expectation that it is misframed.

0:29

Tim Clark: Yeah, assuming it is. We have to assume that it is.

0:29

Junior: Because that alone will invite some dissent.

0:29

Tim Clark: Yes. Right? Not are we?

0:29

Junior: Yeah, are we thinking about this the right way? No, no, we probably are, right? If you use that data, there’s a 37% chance that we are. [10:11:15 - 10:11:15] Tim: That’s right.

0:29

Junior: Thinking about this the wrong way. So how are we misframing this? I think that that can be one way to invite a lot of constructive dissent. So here’s another argument for framing. Fast execution leaves awake. The execution cost getting smaller doesn’t mean that we should execute more necessarily. It’ll often end in that result, but it doesn’t mean that we must. Why? Because there’s a whole bunch of accidental complexity, we call it, that we accrue when we do anything.

0:30

Tim Clark: Anything.

0:30

Junior: And so if previously we could do five things.

0:30

Tim Clark: Even if it’s the right decision. Exactly. Even if previously we could do five things

0:30

Junior: and now we could do 20, it doesn’t mean that we should do 20. We’ll have way more accidental complexity by executing on 20 things than we would if we executed on five. And those consequences will accumulate, they’ll accrue faster because it won’t be linear. We can do all of these things simultaneously. So we have to be very clear up front when we’re framing a problem. I’ll give an example from Leader Factor. A few days ago we were talking about the podcast and we were in a team meeting, it was an all hands meeting and I told the team, I’d like some feedback on the podcast. What I have done historically has been just that, it just stopped there, say, hey, I’d love some feedback on this last episode. Just listen and let me know. And the type of feedback that I got was all over the board based on the person’s preference. And so what I didn’t know is that when I said, give me some feedback on the podcast, a lot of people were hearing, did you like the podcast? And I was getting a lot of feedback about whether they liked the podcast. That’s not the answer I’m looking for.

0:31

Tim Clark: That’s not the question you’re asking. Even irrelevant, yeah.

0:31

Junior: Irrelevant to some degree. Then in this last time, I tried to qualify that and say, here is the criteria for success. Here is the job to be done of the podcast. Here’s what success looks like. Here’s what failure looks like. I want you to watch the podcast as if you were our ICP, the ideal customer profile. Here’s who you are, here are the things that you care about. Now tell me, was the podcast valuable for you and what’s your next likely behavior? That was so much more valuable, spending another two minutes in framing than just leaving it be, right? I mean, first of all, it’s rare sometimes that we invite feedback at all. Two, if we do, we may not be framing the feedback the way that we should. And three, the highest value we could possibly get

0:32

Tim Clark: is by taking just a couple minutes, framing the situation appropriately and then letting people. And so often, Junior, when we frame, we’re injecting confirmation bias into the frame itself. And we’ve already kind of predetermined that we’re going to disregard disconfirming evidence. We’re not interested in it. And we’re just looking for evidence that confirms what we’re doing.

0:32

Junior: Yeah. The last thing that we’ll talk about in terms of challenger safety and decision-making is the rate of change in the environment. If you assume a static environment, you can get away with some of the poor decision-making that we’ve had in times past. You can’t keep pace in an environment like the one we’re in if we don’t have high challenger safety because things are moving so fast. There may be a landscape change, a tooling change that happens two hours after we have a conversation. If that’s not brought up and relitigated, if that’s not part of the decision-making process going forward, we could miss something really consequential down the road. And so people need to feel, they need to feel like they can constantly engage in that way with high intellectual friction disagreement. If they don’t, then we’re not going to be a very smart team. Very true. What would you like to leave people with, Tim, as we wrap up? Any final thoughts?

0:33

Tim Clark: Yes. That innovation is a social process. And the social process depends on your ability to allow your team to disagree. It goes back to your ability to lower the costs of speaking up. And when people feel that they can and you normalize that behavior, now you have the opportunity to create an incubator of innovation.

0:34

Junior: Love it. If you want a team that does decision-making well, lower the cost of speaking up. That is your leadership mandate for the week. Okay, hopefully that was a valuable episode for you all. Certainly was a fun one to treat from our side of the table. And if you found it valuable, we would love your subscription on YouTube. Follow us so that you can get these every single week. Find the downloadable linked as well. And we’ll see you in the next episode. Take care everybody. Bye-bye.