For The Better - Email 7/25

For the Better comes to you bi-weekly with ideas about how and why to build companies focused on human flourishing and stories of the people who are doing it. Other enthusiasms may occasionally appear.

Cultural Conversation vs. ActionThere’s a philosophical term for the category of beliefs we all hold but aren’t conscious of: aliefs. This hard-to-interrogate set includes things like knowing rationally that the characters in a sad movie aren’t real but crying anyway when bad turns of events befall them, or refusing to eat a food you know is perfectly fine just because it’s been shaped into the form of something repugnant (think fudge and dog feces). Aliefs shape our behavior without us even realizing it. Tamar Gendler, who coined the term, tells this story to illustrate the concept: “Gendler once forgot her wallet at a conference, so borrowed cash from a colleague – then reached into her bag to find her wallet, to stash away the money. “Although I believed my wallet was several hundred miles away… I alieved something very different,” she recalls. “The alief had roughly the following content: ‘Bunch of money. Needs to go into a safe place. Activate wallet-retrieval motor routine now’.”Organizational culture is a lot like this in the sense that what we think we believe it is, based on messaging, can be very different from what we really believe it is. For this reason, I like to define culture as “how people behave when they think no one is looking.” To put it more plainly, telling people what the culture is in order to set expectations is limited in its usefulness. Of course companies should reinforce their desired culture through language. Saying nothing leaves a vacuum that will inevitably be filled by behavior ranging from incoherent to awful. Lying about culture is even worse than saying nothing. If a company says it wants to take risks but rewards the opposite, not only is it creating chaos, but it’s giving people impetus to follow in leadership’s hypocritical footsteps. This diagram in Charles Lambdin’s post on why culture isn’t talk is a great illustration of the consequences of failing to realize this. It includes Sean Delaney’s negative ‘leadership shadow’, which signifies the erosion of leadership that comes as a result of giving mixed messages.  

Rather than existing because a company says it does, culture is actually the outcome of incentives and behaviors. If, in the absence of explicit direction, people behave the way the company hopes and expects they will, the culture is sound. Anything else means it isn’t. The only way to change it is to change those two key inputs: incentives and behaviors.Things I’ve Enjoyed LatelyDeaddit This AI-generated version of Reddit caught my eye. It’s a preview of all the bizarre uses of AI yet to come and a fascinating reflection of societal concerns from family relationships to investing for retirement, pet care, recipes, and existential dilemmas.A theory of intelligence that denies teleological purposeThis piece on the From Narrow To General AI blog takes up how working on the question of what the human mind is designed to do and how to measure its success is a key step in figuring out how to best train new AI models. As it points out, “We all feel we know, intuitively, what intelligence is, and yet few of us could describe it in functional terms, or in enough detail to apply it to practical AI projects.” Taiwan is making democracy work again. It's time we paid attentionPolis is a social media platform designed to fosters consensus rather than division and argument. This Wired story about how the Taiwan government used it to do polling designed “to create concrete outputs that the government could turn into new laws” is really inspiring. “They wanted to allow citizens to not only vote on questions posed by the government, but also control what questions were asked in the first place. And they wanted these questions to be based on attitudes held in common across Taiwanese society rather than on its divisions.” Truth is Universal; Robust Detection of Lies in LLMsAI is just like us – it doesn’t always tell the truth. Researchers are devising better ways to separate the falsehoods from the facts, noting that “This behaviour persists even in models trained to be honest.”