Editorial
The Gist
- AI personality isn’t accidental. Companies like Anthropic intentionally shape chatbot traits behind the scenes, revealing them only occasionally.
- Claude’s personality has a design. Anthropic gives Claude a human-like disposition rather than rigid rules, focusing on traits like wit, integrity, and adaptability.
- Personality tuning is deliberate work. Anthropic fine-tunes Claude’s behavior using message-response training that reflects desired values and traits.
Every AI bot’s personality is the product of intentional decisions made within AI research houses, often behind closed doors. We typically only see glimpses of those decisions when things go awry, like in the cases of “white-genocide” Grok or “sycophant” ChatGPT.
But otherwise, the process occurs in the background, influencing the type of conversations we have with the bots without our awareness.
But late last month, Anthropic gave the public a peek into how it tunes Claude’s personality. In a revealing session with reporters at the company’s first developer event in San Francisco, Amanda Askell, a researcher who works on the chatbot’s character, shared how the company is optimizing Claude.
So here’s a look into how Anthropic thinks about its chatbot’s character, and how it actually builds the personality we see:
Table of Contents
- The Entry Point: Global Chatbot Phenomenon?
- The Disposition: Best Friend Ever?
- The Character: Thoughtful Bot
- The Implementation: Feeding Bot With Message Chains
The Entry Point: Global Chatbot Phenomenon?
Claude’s starting point is speaking to people around the world, often without lots of context or full knowledge of their intentions. It’s a difficult problem, Askell said, and requires the bot to take on some human-like qualities to respond.
“Claude’s situation is a bit weird. It has to assist lots of people across the world with lots of different needs,” Askell wrote on a slide. “A starting point might be something like ‘what would an ideal human do if they were in Claude’s situation.’”
Related Article: Contact Centers, Chatbots & Chaos: Solving Today's Biggest Support Issues
The Disposition: Best Friend Ever?
Instead of giving Claude a bunch of rules to follow in specific situations, Anthropic is trying to build a disposition into the bot. The concept of a robot’s disposition is similar to a human’s, where we have inherent qualities that manifest in our behavior. The bot’s disposition should come through in all interactions.
Claude’s disposition isn’t limited to ethical traits such as kindness and generosity; it also encompasses qualities that make someone a good conversationalist or friend, like integrity and wit. When the bot demonstrates these traits in new situations, it’s a sign that the disposition Anthropic is building is working.
The Character: Thoughtful Bot
When you speak with Claude, you get a distinct sense of character as well. “A template here that I like is the idea of a well-liked traveler who can adjust to local customs and the person they're talking to without pandering to them,” Askell said. “They're often very open and thoughtful.”
This also means Anthropic is intentional about whether Claude tries to get us to prolong a conversation, even if it might one day be good for business. “If we think about people who are just trying to engage us, I don't think we often think of those people as good people that we want to hang out with all the time,” Askell said. “We think our friends are good because [they tell us what] we need to hear and don't just try and capture our attention.”
Related Article: Want Deeper Customer Bonds? Start Talking to Their AI
Webinar
Sep
25
CMS Briefing: A Live Look at What’s Next in AI-Driven Platforms
Learn how leading organizations are using AI‑driven tools to publish faster, personalize smarter and stay secure.
Register
Webinar
Sep
25
Call Spoofing Trends Financial Institutions Can't Afford to Ignore
Don't let robocalls sabotage your customer relationships. Learn how to secure your voice channel.
Register
Webinar
On demand
Ready or Not: How Data-First Organizations Are Unlocking Agentforce Potential
Learn how to cut through the noise, activate Agentforce and build a Salesforce AI strategy that actually delivers.
Watch Now
Webinar
On demand
CX Reality Check: What Your Customers Are Really Thinking
Register now and get the insights you need to not just meet rising expectations, but exceed them.
Watch Now
Webinar
On demand
Content Leaders Collective: What a Good CCMS Actually Looks Like
Discover how content leaders transform clunky systems into growth engines — and why your competitors are ahead.
Watch Now
Webinar
On demand
From Hype to High-Impact CX Strategies That Actually Scale
Turn buzzworthy AI and outsourcing trends into measurable CX wins with fresh 2025 data.
Watch Now
The Implementation: Feeding Bot With Message Chains
Much of Claude’s personality is built in the post-training or fine-tuning stage. In this process, Anthropic asks people to write messages where the traits it wants to see would be relevant, and then write responses to those messages that are consistent with the ideal traits. The company then feeds the message chains into the model and asks it to emulate its desired behavior.
How Anthropic gets Claude to model these behaviors could be the subject of another, longer post (and if there’s interest, we can probably do it) but this is largely how the company gets the job done and makes Claude what (or who?) it is today.
Learn how you can join our contributor community.
