Little Guys
I want to post about ‘Little Guys‘ more regularly here on the blog, but to do so I need to establish some design patterns or shorthand to be able to talk about them.
So we are going back to the 90’s.
A couple of months ago I wrote about ‘Little Guys‘ being the emergent AI interface paradigm for Agents. I’ve been writing about ‘Little Computer People‘ and (the dangers) of agents as friends the moment I saw GPT-4o’s voice input too.
Recently, I’ve been consulting and speaking with several teams currently building ambitious agents and AI-integrated software in the Little Guy paradigm. One of the first things I’ve been asking design teams is:
What metaphors were people using for X in the 1990s?
What were they making or doing back then?
Invariably, there’s always a trove of literature and examples that the team can draw on. But! and this has been a particular pet peeve of mine over the last half-decade across the crypto / metaverse / and worlds design space, no one ever goes back and looks at the literature! Younger millennials and zoomers working on this stuff seem to just want to dive in, driven by nostalgia and possibility, without having read a single history book!
Anyway, to understand the current forms of these Little Guys that are being experimented with right now, I’m going to repurpose some of my current client research and I’ll be covering a few key examples of ‘Little Guy‘ design from the 90s and early 2000’s.
Petz: Dogz (1995) & Catz (1996)
The Dogz & Catz series (Petz) from PF.Magic was one of the first commercially successful virtual pets. (I’ll cover Nintendogs, which came a decade later, in a future post).
I grew up with this game, and I have fond memories of playing Catz after dinner on our hearing aid beige Windows 95 machine in the late 90’s. I think it may have even been bundled with the computer?
Aside from the product itself, one of the most interesting parts of this groundbreaking series is how it came about.
The Genesis of Petz
Petz‘ creator was Rob Fulop, veteran Atari designer, and also a designer on the 1992 FMV game: Night Trap. The first interactive movie game on the Sega’s Mega CD.
A game which is cemented in video games history for being the focal point of the 1993 United States Senate hearings on video game violence, where it was accused of promoting gratuitous violence and became a symbol of the perceived ‘moral decay’ in the industry. The outcome of which lead to the creation of the Entertainment Software Rating Board ESRB.
After experiencing the moral panic and the resulting personal fallout, Fulop said to Retro Gamer UK in 2019:
“I didn’t want to make any more games that had any kind of violence in it,” he says. “This made sense to me.”
In a 2022 retrospective interview with gamedeveloper.com he went further:
I fell out with my girlfriend about it, because I thought it was completely bullshit criticism.
But I decided that the next game I made was going to be so cute and so adorable that no one could ever, ever, ever say that — it was, like, sarcastic — what’s the cutest thing I could make? What’s the most, you know, sissy game that I could come out with?
In my opinion, this motivation to create something universally unobjectionable formed the core of the product’s design philosophy.
Playing Petz
Playing Petz hinges on what I call the “Desktop Buddy” paradigm.
Your Little Guy lives inside a window on your desktop, and you interact with it directly using diegetic tools and objects. Click/drag the animals or objects to pick them up and move them around. The key point is the physical and direct manipulation of objects and the agent in the game world. The delightfully retro Petz fan wiki documents these interaction patterns in great detail.
It’s simple, but it really worked. So much so that by 1998, Catz & Dogz, and their sequels had sold over 1.5 million units.
The ‘Desktop Buddy‘ approach, despite being wildly popular with users and generally positive reviews – mostly by people who recognised that there was a large audience seeking different kinds of digital experiences (a trend that would be fully realised a decade later with the rise of casual and mobile gaming).
Petz wasn’t AT ALL well received by a subset of reviewers however; people who were either expecting more or didn’t think it was a game at all. One contemporary reviewer called the the experience “completely pointless” and “amusing for all of three minutes”. The Age, Melbourne wrote in ’96 that “Not since the craze for pet rocks have people so publicly paraded their foolishness.”. Even as recently as last year, WIRED called it “little more than an interactive screensaver”. I suspect that as the modern-desktop-buddy-personified-agents-as-little-guys-paradigm matures beyond the LLM chat window in our current era, we are going to see similar objections and sentiments.
However all along PF Magic’s ambition behind Petz was to create “interactive, real-time autonomous characters”, not full-blown simulations.
The behaviours of the Petz: getting hungry, howling for attention, staring at your mouse cursor as it moved across the screen etc brought just enough personality to engender what I call aliveness.
See these posts for more on Aliveness in agent design:
They are also cute as hell, which goes a long way towards personification.
The AI in Petz wasn’t sophisticated by today’s standards, but the agents did have a basic reinforcement learning loop. You had a few tools: Treatz to reward, a spray bottle to scold. But the gaze-following was very effective. It makes you feel like your pet saw and noticed you.
That one feature probably did more heavy lifting towards ‘aliveness’ than anything else in the whole game.
Gaze-Following
Gaze-following is a pattern I’m increasingly seeing in modern AI agents. Like the ‘little guy’ computer which I’ve written about before.
We’re also seeing it in electric vehicles from Chinese brands. Both NIO’s NOMI (a) and BYD’s Qin (b) have dashboard agents. Little guys that live on the cars dashboard, and respond using gaze tracking when you’re in the car.
If you want to ‘drive’ deeper, I highly recommend this 2022 paper: Multimodal Warnings Design for In-Vehicle Robots under Driving Safety Scenarios.
The research on multimodal warning design for in-vehicle robots, especially the multimodal warning with combined visual modality and auditory modality, is of critical importance for human–robot driving safety during human–robot co-driving. This includes when each modal warning is displayed, what is included in each modal warning, and how robots express their facial expression and give voice warnings. This paper first discusses the content and coordination of robot expressions and speech based on SAT theory, then discusses robot facial-expression valence and speech rate. Then, different multimodal designs are experimentally evaluated using a robot equipped with a facial screen and a speaker in a scenario where the driver makes a mistake (speeding).
It’s also worth exploring this two page paper from PF.Magic about the design of Petz, produced for the AGENTS ’98: Proceedings of the second international conference on Autonomous agents conference where they explore Interface design, pet behaviour modelling, and user interpretation feedback and more.
Community and Legacy
Another important aspect of the Petz series we must consider in its history as a ‘Little Guy‘ is the community that grew up around it.
Despite it’s release very early in the Internet era, PF.Magic added a camera tool to the game that let players take .bmp screenshots of their pets and upload them to their official website. A feature that gently nudged young people toward the web. Often the Petz website being one of their first online experiences and destinations. It certainly was one of mine.
Players could enter photo contests like “Mutt Mugs” or share stories in the “Dogz Tails” forum. It was early, and very savvy online community building for the time.
Moreover PF.Magic also took a remarkably open stance on modding. Not only could players edit the game files, but the company quietly encouraged it, sometimes even providing how-tos.
The results were wildly successful. Users reverse-engineered the system and began creating new breeds (“breedz”), toys, playscenes, and even dev tools. What started as a game with five dog types became an endlessly expandable platform. Years before user-generated content became a design cliché.
This in my view this prefigures everything from Minecraft to the Steam Workshop. If anyone knows of earlier modding community of this scale please leave a comment! ↓
Here’s some links to Wayback Machine archived sites with eye melting aesthetics if you feel like exploring what this community was like for yourself.
There’s a story here I think, the sort of thing Katherine Dee might write/be interested in.
Low-Pressure Companionship
Lastly, one of the most important aspects of Petz design philosophy (and what sets the Petz series apart from contemporaries like the Tamagotchi) is the total absence of stakes.
Your pet couldn’t die. It didn’t get sick, or shit on your screen.
There was no risk of failure, no punishment for forgetting about it. It was designed to be emotionally engaging without being demanding. Less simulation of care (a very important concept in the current AI design space in Itself), more low-pressure companionship. Which we also see continued though in modern ‘Toy Box’ games like Animal Crossing, and probably originated with 1985’s Little Computer People on the Commodore 64 which inspired The Sims.
‘Low-Pressure Companionship’ is a phrase I think applies quite well to in-car robot UIs too. Especially those that signal feedback with a wink or a smile. There is a space and place for this in the current design paradigm, at Develop this year I saw several LLM powered ‘Terrariums’—wind-up worlds with agents inside of them.
But as agents become more interactive, with speech and personality, and emotional bonding as part of its design, I remain concerned. The friends are virtual, but the feelings are real. As I wrote last year
Whilst the mainstream front in the AI Culture War is currently focused on copyright and ‘is it art’, the rest of the AI industry is full steam ahead making and designing virtual agents. The implications resulting from their deployment is really something I really think we should be talking about.
The lack of lifecycle design in the current crop of personified AI agents that can trace their way back to Petz, is bad design for these kind of modern agents, and is something I’ve written about at length, so I won’t repeat myself.
In the next post, I think I’ll be writing about Clippy (1997) and Microsoft Bob (1995). I also plan to write about the Sony AIBO robot.
What else should I include? Let me know in the comments ↓
With the design of agents and avatars of course, comes the need to design worlds to put them in. Why not check out my 30k word essay collection on World Running?
Leave a Reply