At the same time as synthetic intelligence (AI) turns into embedded into each material our each day lives — from language translation to digital private assistants — it continues to be a divisive subject. As its attain expands, Virginia Tech researchers are searching for to know which sections of society may be extra receptive to AI and which sections could also be extra averse to it.
Within the not too long ago printed analysis “Partisan Media Sentiment Towards Synthetic Intelligence,” authors from the Virginia Tech Pamplin School of Enterprise – Angela Yi, Shreyans Goenka, and Mario Pandelaere – examined the various reactions to AI by analyzing partisan media sentiment. Their work was printed within the journal Social Psychological and Character Science.
The researchers discovered that articles from liberal-leaning media have a extra destructive sentiment towards AI than articles from conservative media. In different phrases, liberal-leaning media are usually extra against AI than conservative-leaning media.
This opposition will be attributed to, in response to the findings, liberal-leaning media being extra involved with AI magnifying social biases in society, akin to racial, gender, and earnings disparities, than conservative-leaning media. The researchers additionally examined how media sentiment towards AI modified after George Floyd’s demise.
“Since Floyd’s demise ignited a nationwide dialog about social biases in society, his demise heightened social bias issues within the media,” stated Yi, a Ph.D. pupil within the advertising and marketing division. “This, in flip, resulted within the media changing into much more destructive in direction of AI of their storytelling.”
Implications for policymakers and past
Based on Goenka and Yi, their findings might have necessary implications for future political discussions round AI. Becaise media sentiment can function an indicator of public sentiment which, in flip, can affect policymakers’ stances, the partisan media variations noticed might subsequently result in variations in public opinion towards AI.
“Media sentiment is a strong driver of public opinion, and oftentimes policymakers look towards the media to foretell public sentiment on contentious points,” stated Yi. “Maybe the subsequent step in our analysis is to see how social media conversations surrounding AI change as a operate of the partisan variations we see in our paper.”
How the info was collected
To look at partisan media sentiment towards AI, the researchers compiled a set of articles written about AI from a number of media shops. The partisan sentiment for every outlet used was decided by utilizing the scores discovered on the Media Bias Score Chart from AllSides, an organization that measures the perceived political bias of content material on on-line written information shops. A mixture of liberal-leaning shops, akin to The New York Instances and The Washington Publish, and extra conservative-leaning shops, akin to The Wall Road Journal and the New York Publish, have been sourced.
From there, the researchers downloaded articles from the chosen shops based mostly on sure standards, together with the utilization of particular key phrases, akin to “algorithm” or “synthetic intelligence,” in addition to a date vary from Might 2019 via Might 2021.
With a dataset of over 7,500 articles, they carried out an emotional tone evaluation on every story utilizing an automatic textual content evaluation device. By way of this device, they have been capable of seize the emotional tone of every article, which is calculated by the distinction between the proportion of constructive emotion phrases and the proportion of destructive emotion phrases in a textual content. This distinction is then standardized on a scale of 0 to 100 to supply the emotional tone measure.
Goenka, assistant professor of promoting, pressured that this analysis is descriptive moderately than prescriptive, and no stance is being taken as to the precise technique to talk about AI.
“We aren’t stating whether or not the liberal media is performing optimally, or the conservative media is performing optimally,” he stated. “We’re simply exhibiting that these variations exist within the media sentiment and that these variations are necessary to quantify, see, and perceive.”