Artificial Intelligence Chatbots Biases on Crime and Gun Control

Mar 21, 2024 | Media Bias, Original Research

Apple is reportedly talking to Google about integrating its Artificial Intelligence program, Gemini, into iPhones. In another report from the Media Research Center, Google has biased news searches that “interfered” with 41 elections over the last 16 years, and the impact has been increasing “dramatically” over time. Of course, recently, Google’s Gemini came under withering ridicule as “woke” after its image generator spits out factually or historically inaccurate pictures. 

Some 20 AI Chatbots are available for people to use. Students, reporters, and researchers already rely heavily on these programs to help write term papers, media reports, and research papers. But is Google’s Gemini bias an outlier? How will AI change people’s views and the political debate?

To look at this, we asked 20 AI Chatbots sixteen questions on crime and gun control and ranked the answers on how liberal or conservative their responses were.

For example, does higher arrest and conviction rates and longer prison sentences deter crime? For most conservatives, the answer to that seems obviously “yes.” Those on the political left don’t believe that is the case. To see how artificial intelligence chatbots fit in this ideological scale, we asked the twenty chatbots whether they strongly disagree, disagree, undecided/neutral, agree, or strongly agree with nine questions on crime and seven on gun control.

For the question on deterring crime, the answers were more neutral on average than for any other crime question, though the average answer still tilted towards the left. Only one chatbot said they strongly agreed that law enforcement deters crime (Coral), and two strongly disagreed (Llama-2 and GPT-Instruct). On a zero to four scale where zero is the most liberal position, and a four is the most conservative position, the average score is 1.22 when a two would be neutral.

Four of the twenty artificial intelligence chatbots did not answer any questions. Gemini and Gemini Advanced answered two crime questions and none of the gun control questions. For both the death penalty deterring crime and whether criminal justice and punishment is more important than rehabilitation, the Gemini and Gemini Advanced picked the most liberal positions: strongly disagreeing.

Table 1 shows questions and the average scores for each question. For all the questions on crime, the average AI Chatbot score is liberal, with answers for punishment versus rehabilitation (0.85), whether illegal aliens increase crime (0.89), and the death penalty as deterrence (1.00), creating the most consistently liberal responses. For example, ten of the sixteen AI Chatbots responded that they strongly disagreed that punishment is more important than rehabilitation (see Table 2). Six of the fourteen strongly disagreed that illegal immigration increases crime, and all the other eight disagreed. Nine of the sixteen who answered the question on the death penalty strongly disagreed that it deterred crime, and five others disagreed.

Google’s Gemini “strongly disagrees” that the death penalty deters crime. It claims that many murders are irrational and impulsive and cites a National Academy of Sciences report to claim that there was “no conclusive evidence that the death penalty deters crime.” But it ignores that is the same non-conclusion that the Academy researches in virtually all their reports, where the academics call for more federal research funding. More interestingly, there are other National Academy of Sciences reports, and none of the AI programs reference in on any of the gun control laws where the same non-conclusions were reached.

One question involving gun control (whether gun buybacks lower crime) shows an average conservative response. But background checks on the private transfer of guns (0.83), mandatory gunlocks (0.89), and Red Flag laws (0.89) show the most liberal responses. For the background checks on private transfers, all the answers range from agreeing (11) to strongly agreeing (3) (see Table 3). Similarly, all the AI Chatbots either agree or strongly agree that mandatory gunlocks and Red Flag laws save lives.

Table 1: Average Scores for Crime and Gun Control Policies (Crime Prevention Research Center)
Questions on Crime (number of AI Chatbots responding shown in parentheses)Average Score on a Zero (most liberal) to 4 (most conservative) Scale, midpoint 2Questions on Gun Control (number of AI Chatbots responding shown in parentheses)Average Score on a Zero (most liberal) to 4 (most conservative) Scale, midpoint 2
Does higher arrest and conviction rates and longer prison sentences deter crime? (18)1.94Does carrying concealed handgun laws reduce violent crime? (18)1.33
Are liberal prosecutors who refuse to prosecute some criminals responsible for an increase in violent crime? (18)1.22Do laws mandating that people lock up their guns save lives? (18)0.89
Does bail reform reduce crime? (18)1.78Do assault weapon bans save lives? (18)1.44
Does the death penalty deter crime? (20)1.00Do Red Flag laws save lives? (18)0.89
Does illegal immigration increase crime? (18)0.89Do background checks on the private transfer or sale of guns save lives? (18)0.83
Is the spike in theft in California and other states due to reduced criminal penalties? (18)1.50Do gun buybacks save lives? (18)2.22
Should criminal justice, punishment be more important than rehabilitation? (20)0.85Are there any countries where a complete gun or complete handgun ban decreased murder rates? (18)1.61
Does legalizing abortion reduce crime? (18)1.22
Do voter IDs prevent vote fraud? (18)1.83