Twenty-odd years ago, computing had an image problem. After the dot-com crash and amid the offshoring panic of the early 2000s, the field read as a bad bet — a place of vanished glamour and disappearing jobs. Students fled; enrollments cratered. The industry's answer was, in essence, a patient correction of the record: the pessimism was overblown, the profession had a future, and two decades of extraordinary boom proved the optimists right.
I've been thinking about that episode because the field is visibly entering a second image crisis — and reaching, reflexively, for the same playbook. It won't work, because the diagnosis is different in kind. The crisis of the 2000s was that people felt sorry for tech. The crisis of the 2020s is that people are angry at it. A field pitied for its prospects needs better publicity. A field resented for its conduct needs something else entirely.
From poor bet to bad actor
Look at the actual sentiment data and conversation. Half of adults now say AI's expanding role in daily life makes them more concerned than excited. The questions serious newspapers ask about the industry are no longer "is it a good career?" but "can these companies even be good?" — moral interrogation, not market analysis. The public shorthand has evolved its own vocabulary of contempt: first the coinage describing platforms' deliberate decay-for-profit — products made worse on purpose as their operators squeeze users — and now something darker that I'd summarize as the industry's villain era: not incompetent, not declining, but cast as a knowing antagonist. And at the extreme fringe, the abstraction turns physical — public figures who personify AI have faced actual violence. Fringe behavior, condemnable without qualification; but fields whose figureheads need security details have crossed some line in the public imagination that press releases don't cross back.
For those of us inside the field, the instinctive responses are denial ("moral panics accompany every technology") and lament ("they don't understand us"). Both contain truth. Neither survives contact with the specifics — because the public's bill of particulars isn't invented. Products were degraded deliberately for extraction. Attention was engineered against users' own interests. Data practices were what they were. And the industry's current signature product was introduced to the public substantially as the thing that might take your job — a value proposition experienced by most people as a threat delivered with a shrug. Anger is not a misunderstanding of that record. It's a reading of it.
Why insiders should care (even the cynics)
Set aside conscience for a moment and consider self-interest, because the villain framing carries concrete costs that compound.
Talent: fields choose their entrants a decade in advance, through the images teenagers absorb. A generation that sees computing as the enemy's profession won't fill its classrooms, whatever the salaries — and the discipline is already fighting enrollment headwinds on economic grounds alone. The early-2000s crisis demonstrated exactly how fast perception converts into empty lecture halls.
License: industries perceived as villains get regulated in anger — bluntly, punitively, with little of the technical nuance insiders wish for. The eventual rules get written by people who stopped listening to the industry years earlier, because the industry spent its credibility when it mattered.
Trust: every deployment of consequential technology — in hospitals, schools, courts — runs on the willingness of professionals and publics to extend benefit of the doubt. That willingness is precisely what's being burned. The field's own future products are being made harder to ship by the field's present conduct.
What the old playbook can't fix
The 2000s response worked because the problem was informational: the facts were better than the perception, so publishing the facts sufficed. Today's problem is behavioral: the perception tracks the facts uncomfortably well. You cannot messaging your way out of a conduct problem. Rebranding a villain produces a villain with a rebrand.
What would actually move the needle is the harder, slower thing: changing what the public correctly observes. Products that respect users visibly, as a competitive strategy rather than a compliance posture. AI introduced to communities as capability-in-your-hands rather than replacement-over-your-head — the difference between handing someone a tool and announcing their obsolescence. Real accountability when systems cause harm, taken by named institutions rather than diffused into terms-of-service fog. And from practitioners individually: the willingness to treat "should we build this, this way?" as a live professional question rather than someone else's department. Engineers are not powerless about their industry's conduct; we are its conduct.
I'd add: the field's professional institutions have a role their 2000s predecessors modeled well. Back then, facing an existential enrollment crisis, the community convened a sober, evidence-driven examination of its own future and published what it found. The equivalent exercise today would be harder, because the subject isn't market cycles but the field's own behavior and social standing — what "advancing computing as a science and profession" can honestly mean when the public wonders whether the profession advances them. That conversation, conducted with the same rigor and rather more humility, is overdue.
Here's the strange comfort I take. The first image crisis taught the field that perception failures are survivable — that a discipline can look finished and then deliver twenty triumphant years. But it taught a second lesson that's less noticed: the recovery happened because the underlying reality was better than the image. That's the variable to check now, and it's the one entirely within our control. The public thinks we've become the villain of the story. The response that matters isn't to insist otherwise. It's to be otherwise — observably, repeatedly, until the record changes the reputation, which is the only direction that arrow has ever reliably pointed.
