Max Greenwood
As artificial intelligence tackles the business of politics, some pollers are turning to new technologies for new insights.
AI-assisted votes are still in the early stages. However, a small number of companies are looking for ways to tackle everything from creating research questions to segmenting and analyzing huge amounts of data. In some cases, companies like OpenResearch, a nonprofit funded by OpenAI, are experimenting with AI chatbots as part of the polling process.
In a recent interview with Campaigns and electionsAmir Kampurwala, co-founder of the outward intelligence of quantitative research platforms and Harris poll alumnus, has argued that it can be used to eliminate technology, or at least mitigated, to investigate clients' clients in traditional voting and generate insights.
Of course, there is a risk, Kampulwala said. Bad actors can use AI bots and agents to distort data and erode confidence in the vote as an institution, he said. Also, voting businesses are susceptible to the same pressures that AI imposes on other industries, and some people are likely to lose their jobs as AI plays a bigger role, Kanpurwala said.
However, he also argued that there was reason to be excited about the AI-assisted vote.
The following interviews have been compiled for length and clarity.
C&E: Set our stage a bit. Where are we now with AI and polling? What does that space look like?
Kanpurwala: We are in this very early place about how AI affects the world of market research. There is optimism of it, there is pessimism of it, and I start with pessimism.
Voting is extremely important. It determines how we talk about things and how we perceive the world around us. It is used to understand how others think about the problem and whether you are on the right side or wrong in that problem. In politics, it helps us understand who is ahead of us, and it can really help or hurt the campaign.
AI can throw a wrench into it to some extent. If you are collecting data, if some of it is human and some of it comes from AI bots or agents, you can question a lot of the data being collected. And I think we're very early on from the pessimistic stage about how AI cascades through the data being collected.
However, I think the AI section is extremely beneficial for all parts of the process in terms of how the research is conducted and the widespread information. And people don't realize how complicated the process is. You start with the question: I want to understand better how voters perceive these different candidates, or how consumers think about my company. However, there are many steps to how to actually answer that question from that question. You need to write a survey and make sure those questions are properly expressed. So, it can be used to help you create questions that minimize bias using AI.
So, is that question of how the investigation reaches people? Outward Intelligence conducts all your research online. And it's either someone's desktop, tablet, or phone. But we need to be able to incorporate that written investigation and design into the programmed ones. Use AI to proceed from here. It makes sure that the list of questions is logically connected.
We also use AI to collect very representative data. Looking at some of the campaign's previous voting cycles, people will say that representation is off – the number of Republicans and liberals and conservatives in this study is not accurate.
C&E: How do people respond to this idea that AI asks questions, questions, and leads conversations?
Kanpurwala: I think our goal is to make it as seamless and intuitive as possible. So, for our company, OI, when people are doing research, they do it in their browsers. We are responsible for ensuring our research experience does not feel outdated and like drag.
It also needs to be intuitive. So the questions are asked in a way that appears to be repetitive, redundant, misleading or unbiased. AI really needs to disappear into the background and allow people to communicate their ideas. To that extent, it is to navigate the right questions to the right respondents and not flood the same repetitive topics.
Political research may have different questions about favors, or different policies or voting measures. We know the zip code you are in. Whether you participated in a previous election or not, we know by matching your data to your voting file. Therefore, AI should be able to coordinate all this information about respondents and ask questions related to them in real time.
They should never know that there is an AI that is deciding on this. The key is that the questions they answer relate to them and that they care about.
C&E: To what extent is this technology trying to affect the industry and the people working within it?
Kanpurwala: We talk to people who work in these companies – every day with many people. And you will see that there is unemployment happening in this field right now. And that's because AI is easier than ever at every stage of the research process. Anecdotesly, I am sadly seeing that there is unemployment happening. This is the period of stages.
C&E: I just said that when it comes to AI in opinion polls, there are reasons for optimism and reasons for pessimism. Are you most concerned about this technology?
Kanpurwala: Start with what I worry most. It's the ability of a bad actor to influence the data we see, both on a state-based agenda and on a state-based agenda.
We are reflecting this, a social fabric, that is the poll. And I think we, as a country, have faith in that data, probably at the total level, not for individual polls. So what happens when people try to manipulate it? What can you believe? How can you be completely certain that the data you have is not corrupted in any way or fashion?
We're building a lot of resources and tools that can detect AI-based agents while doing research. But it's a cat and mouse game and people are going to invest more and more in this field. And it may not necessarily be because of some completely creepy things… but even if it's not a slut of maliciousness, it may have the same effect as long as it reduces people's trust in their votes. They are what keeps me up at night.
As for how optimistic I am about AI, this service (polling) can be offered at a much more accessible price range, much faster. I did research for my clients in the past when they wanted to see how their company perceptions change every day. And it could take as much as 90 days for businesses to decide on old data before AI and some of these latest tools emerge. So, using these tools, you can back up your data to people in just one day, not 90 days.
I think AI will help decision makers understand what is going on and respond to it in a very short amount of time and with just a small portion of the price they have paid historically.
