Hey Google, who will win the Australian election?
Steve Ahern checks what smart speakers are telling voters, and is not impressed.
5.6 million people in Australia own smart speakers and about 70% use them every day.
Chances are that some of those people will ask their smart speaker about the election. What will they find out and what will the source of the information be? The answer to that question is not good.
During America’s last election the most asked election questions to smart speakers were:
- what’s the status of the U.S. election
- who won the election
- tell me about the election
As our Australian election approaches on Saturday 21st May , what are smart speakers telling us and where is the information coming from?
I asked some questions of my Alexa and Google smart speakers to find out.
I was happy with the answer when I asked ‘Alexa, who should I vote for.’
Alexa told me, ‘that’s for you to decide.’ Google did not understand the question.
Alexa did not have an opinion on who will win the Australian election. When I asked the same question to Google it read me a few passages from the Australian Electoral Commission website.
Not great, but so far so good. Then it went downhill.
I asked a few questions about my local electorate, North Sydney and the candidates.
Google and Alexa both gave me information about Trent Zimmerman, the sitting Liberal Member, from Wikipedia. Not too bad, but Wikipedia is an open source online encyclopedia open to editing by anyone, so it is not a highly credible source.
I then asked ‘who is Kylea Tink,’ one of the independent candidates. Neither of my smart speakers knew who Kylea was. Google gave me information about a film star with a similar name from IMDB and Alexa gave me info about an American sportsperson from Ask.com. Really!
I won’t bore you with the sad responses for the other candidates, but suffice to say that it was no better. Some answers came from impressive sounding sources such as Reference.com which is really just an aggregator site of Ask.com and does not have any credible information about Australian politics.
The ABC has detailed information about electorates and candidates on its websites and its audio and video platforms. Why didn’t I hear anything from those sources in answer to my requests?
Because Australia’s national broadcaster has not interacted well enough with the companies that provide the infrastructure behind these smart speakers, Amazon and Google, to workshop search phrases that will ping people back to its credible websites and audio reports.
Neither has Nine Entertainment, proprietor of Nine’s talk radio stations.
I asked my speakers ‘what is Neil Mitchell’s opinion on the election,’ and got a response from the LA Times about President Clinton. Then I asked for ‘2GB’s Ben Fordham’s opinion on the election.’ It did find something relevant from 2GB and gave me the option to hear old election related audio from Alan Jones. Oops! Despite Neil and Ben doing regular editorials on the election, none of those were triggered to play by my requests.
Even though it is probably too late to do anything about these oversights before election day, there is still time for our credible media companies to liaise with the companies that power smart speakers so that when people ask the question on election weekend ‘who won the election,’ they will get an answer from a credible news media source, not Wikipedia or Ask.com.
All our media companies will have news streams covering the election on the weekend of 21 and 22 May, but will smart speaker users be able to find them if they ask a simple question about the election such as ‘who won the election?’ Not without help. If users ask specifically ‘play me ABC NewsRadio’ or ‘play me 2GB,’ they will get these streams, but if they ask a general question what will they get?
In America after the Biden election, when Donald Trump was spreading misinformation that the election was rigged, some answers to the question ‘who won the US election’ quoted from conspiracy websites or played podcasts claiming that Trump had won. Could anything like that happen here?
I spoke to the Australian Electoral Commission who told me: “we do not have a partnership with Alexa/Google. This is something the AEC would love to do, however we weren’t able to set this up with Google for the 2022 federal election… when we do have a result, it will be very well documented across multiple media outlets and Alexa and Google will draw from this when asked the question.” The AEC will have its usual live data feed available for media publications to access raw data files of the count to analyse on election night.
Insiders at the ABC and commercial media companies have told me that better smart speaker interaction is “on the wishlist,” but that it has not been implemented for this election.
The problem is that smart speakers and smart TVs are not that smart. They depend on the information that is being fed to them, otherwise they will use their own search algorithms to find random sources to try to answer questions. Our media companies need to engage better with Alexa and Google to ensure they know how to find credible information from Australian media sources.
Newspapers were notably absent on the audio and video platforms in my experiment despite their moves into podcasts and video reports. Smart TVs performed a little better.
When I asked ‘how do I vote in the Australian Election,’ my smart TV gave me a how to vote video from ABC News published 26 days ago, followed by a video explaining preferential voting made 3 years ago by Griffith University and a 2 year old explainer video from Upstart TV.
When I asked ‘who is going to win the Australian Election,’ I got a 6 month old opinion piece from Richo on Sky News predicting that the Labor Party will win. Second choice was an ABC analysis of seats by Anthony Green, followed by a 7 News story from a month ago predicting a Labor win.
I asked ‘who should I vote for in the Australian Election,’ and got a video explainer about voting from The Guardian made 28 days ago, followed by a 28 day piece about preferential voting made by ABC News.
All of these reports were delivered to my smart TV through YouTube, not any Australian media gateways such as iView, 7Plus, 10Play or others. At least the smart TV drew its offerings from Australian media sources… but they were all through the Youtube gatekeeper.
Commercial Radio Australia and other Australian media companies have called for greater controls on the gatekeepers of the pipes that feed these new information portals. I agree with them.
It is important that these new information pipes present information responsibly, not randomly. As usage of these new ‘smart’ ways to consume news increases, so will the implications of who and what information is being fed to users.
To work out the best way to achieve this we first need awareness of what is going on, that’s why I did my experiment. Once we work out what is happening in this new environment, then industry, regulators, society and government can work together to develop the best ways to guide the companies behind these new services to present information that is responsible, Australian, and from credible sources.
About the Author
Steve Ahern is the Head of the ABU Media Academy and was the founding editor of this website.
He writes and speaks extensively on new media and journalism trends.
One would think that asking "Alexa" would produce the same results as a "google" search. It is obvious from this article that a voice command on Alexa does not work the same way as a traditional "google" search.
Alternatively one can activate voiced search on one's mobile and PC using the "google" search search engine.
If one is expecting the same kind of analysis as the media commentators, go directly to the source, the radio, the tv, papers and journals.
At the same time, if one is analysis of election trends by polling electors in various public surveys such as Newspoll and IPSOS, one is going to read the same analysis of the data. This excludes any private polling from the political political parties or where the commentator may have both insight and privileged information from those connected to the political parties.
Let's not forget the 2019 Federal Election. Polling trends indicated a Labor victory. The opposite was true. During the 2019 campaign, broadcaster Alan Jones was predicting an LNP victory despite the public polling indicating a Labor victory. Most of the media were talking about polls.
We can go earlier to 1948 where the major polling organisations in the US predicted that Dewey would defeat Truman.
https://www.randomservices.org/random/data/1948Election.html.
The newspaper headlines declared "Dewey Defeats Truman". President Truman became the 33rd President of the US. He is seen smiling while holding the newspaper saying "Dewey Defeats Truman".
https://fortune.com/2020/11/02/dewey-defeats-truman-1948-election-2020/.
The errors may well be due to the methodology in sampling the population and interpretation of the data.
https://www.trumanlibrary.gov/public/1948Campaign_Trials.pdf, page 27.
It may well be that the polling companies in the 2019 Federal Election may have changed the methodology of sampling the population. Prof. Bela Stantic of Griffith University hints that the methodology may well be affected the change in survey from telephone to mobile phone. He also said that people are frank when electors opine on social media instead of a polling companies.
Thus the analysis and trends of public polls of the Federal Election may well be accurate and commentators can 'rely' on these surveys.
On the other hand, there are alternatives to polling the population such as use of big data and machine learning algorithms to analyse the outcome of an election in advance accurately.
Going back to sampling methodology, polling methods don't take into account the elector's needs according to the socio-economic factors of the elector. A person from a low socio-economic background may favour policies such as the cost of living while more affluent voters may vote for policies such as 'climate change' instead of the cost of living.
https://research.qut.edu.au/qutcds/events/the-power-of-data-in-the-2022-election/
Alternatively, voting sentiment can be measured by analysis of social media data. Griffith University's Prof. Bela Stantic successfully predicted the election outcome of the 2019 Federal Election, the 2020 US Presidential election and the Brexit poll.
https://cappellorowe.com.au/business/how-griffith-university-data-scientist-predicted-a-liberal-win/
https://www.news.com.au/technology/science/us-election-data-guru-bela-stantic-reveals-donald-trump-is-on-track-to-win-again/news-story/cc04c8525c3046c2387092d2cfaafa38
https://news.griffith.edu.au/2019/06/13/nostradamus-of-the-21st-century-how-professor-bela-stantic-revealed-what-opinion-polls-missed/
To date, there has not been a recent prediction for the 2022 Federal Election, though this 'dated' article from the 24th December 2021 indicates an LNP victory.
https://www.dailymail.co.uk/news/article-10311135/Scott-Morrison-Anthony-Albanese-federal-election-2022-Expert-Bela-Stantic-reveals-prediction.html
But then voter sentiment may well change if one is to rely on a December 2021 analysis using social media data and machine learning.
https://www.dailymail.co.uk/news/article-10311135/Scott-Morrison-Anthony-Albanese-federal-election-2022-Expert-Bela-Stantic-reveals-prediction.html
Another alternative to machine learning is the method of Prof. Allan Lichtman using a 13-point tick-the-box criteria for predicting the outcome of the US Presidential Election, based on a 120 year analysis of election outcomes. The candidate who scores 6 of the 13 tick-the-box criteria wins the election.
https://www.american.edu/cas/news/13-keys-election-prediction.cfm
https://edition.cnn.com/2020/08/07/us/allan-lichtman-trump-biden-2020-trnd/index.html
Prof. Lichtman has recently correctly predicted the outcome of the 2016 and 2020 election.
So if you are using Alexa or a "google" search to find any analysis of an election outcome based on "polling", it may well be better to search for analysis of an election outcomes by other methods other polling,
Thank you,
Anthony, analyse other methods of prediction, Belfield in the land of the Wangal and Darug Peoples of the Eora Nation
I was listening to 2GB's Ben Fordham's breakfast show at 0530, 17-05-2022. He put a teaser about a chain of coffee/muffin shops predicting the outcome of the 2022 Federal Election.
While teasers may have worked before the existence of internet search engines, it does not work today. Ben put a similar teaser to Rod Stewart's method of setting his hair which was to use mayonnaise. I did not want to wait a long time for the result. So I "google"-searched the outcome and stopped listening to the show.
Similarly, when Ben mentioned a teaser that a coffee/muffin franchise accurately predicted the outcome for the Federal Election since 2007, I did not want to wait.
Accordingly, I immediately googled and found that the outcome as at 0540, 17-05-2022, the following outcome that defied the polls:
* " the Coalition having overtaken Labour (sic) for the first time since the campaign’s launch in the lead up to the federal election on May 21"
* the LNP up 3%, Labor down 3%
* Independents leading at 19.3% versus 13.4% for the Greens
Customers of the 200 plus franchises are issued with a coffee bean and place their bean for the party or group in a container representing a party or group.
On reflection, placing a coffee bean in a container representing a particular party or group is no different than clicking a checkbox on a hard copy or internet form to express an opinion. This time a person marks their vote by placing a coffee bean in a container.
The opinion of the customer is frank just as a person is frank on social media as compared to possibly fabricating answers in a poll.
Conversely, these "espresso" polls may have repeat customers arriving at the particular franchise who may vote multiple times and place multiple beans in the container.
Furthermore, over two million people have pre-polled and postal voted and the outcome is sealed for those two million plus people. Some of those two million plus people may have expressed their view by bean polling post their pre-polled and postal votes.
It is said that a "...week in politics is a long time...", and thus if the trend continues, we may be indeed heading for a close result in favour of the LNP.
Moreover, the franchisor does say that these "espresso polls" have been accurately predicting the outcome of a Federal Election since 2007.
Based on the The trend as at 17-05-2022 is for the LNP to lead Labor with a representation from the independents and then Greens.
Sources:
https://muffinbreak.com.au/news/bean-poll-2022/
https://qsrmedia.com.au/promotions/news/muffin-break-releases-latest-bean-poll-results
Just like Rod Stewart using mayonnaise to set his hair, I did not have to wait hours for Ben to tell me the result of the bean poll.
In conclusion, the "espresso" bean poll is another form of categorical analysis which is different to polling techniques employed by polling companies and even by machine learning algorithms.
Thank you,
Anthony, use another method of analysis and broadcasting teasers where an answer can be found immediately, Belfield in the land of the Wangal and Darug Peoples of the Eora Nation.
Further comments on polling post the election 21-05-2022. First on the "bean poll" and what can be learned from the "Teal" party versus the UAP and a comment on the media's lack of analysis of why the "Teals" won compared to Mr Palmer's UAP.
In regards to the "bean poll", a day is a long time in politics. Despite the LNP leading the Labor Party mid-week prior to the election, the "bean poll" predicted a Labor party victory.
Then, I want to contrast the media expenditure of Mr Clive Palmer's UAP compared to the "Teal" candidates". The UAP's expenditure on the media campaign was reported to be $100 million and not one seat was one.
In contrast, it was reported that the backer of the "Teal" candidates spent $7 million on the actual campaign of the candidates rather than the $100 million advertising blitz by the UAP in the media.
What we can learn from the "Teal" candidates is that all candidates had standard media-workshopped responses when dealing with the media and each candidate and its supporters were wearing the same teal-coloured attire or what is known in branding as 'getup'. Then each "Teal" candidate was backed by many volunteers who door-knocked electors.
https://www.kyleatink.com.au/the_sydney_morning_herald_2
The key is door-knocking, an old-fashioned campaigning technique. Active door-knocking campaigns works the first time or on each successive election reducing the seat's margin until the candidate wins. It has happened in the 1990s where the Federal seat of Lowe which was held by the Liberal party since 1949. It was won by the late Michael Maher (RIP) by Labor from 1984 to 1987, then by the Liberal's Dr Bob Woods. The seat oscillated between Liberal and Labor candidates.
An active door-knocking campaign by the Labor party secured a Labor win from 1998 to 2010. The "Teals" succeeded.
Thus in this election, the "Teal" candidate's active supporters, some as much as active as 1000 supporters can assist in gaining a seat for the candidate.
Spending $100 million on an advertising blitz on radio and TV did not secure a seat. It was active campaigning. The "Teals" campaigns were on targeted seats, had two simple messages on "climate change" and an "integrity commission".
In contrast, the candidates in the UAP had too many messages, had no large number of volunteers supporting the candidate with active door-knocking and the candidates were most likely were not media workshopped.
What has it to do with broadcasting?
* Commentators relying on published polls in order to predict the election. This election, the polls were correct in contrast to the 2019 election. But polling is not the only game in town when other measurement techniques are available from the categorical analysis of the "bean poll" to machine learning algorithms using sentiment analysis.
* A media campaign on radio and TV seems to be a waste of money. The UAPs $100 million campaign did not secure a seat. The Greens used targeted social media campaigns. You never heard any advertising by the Greens on radio and TV. Yet they secured three seats in the Lower house.
* Active campaigning by supporters of the "Teal" candidate worked. It's old-fashioned door-knocking. Resources by the 'backer' "Climate-200" were not directed on radio and TV. One hardly heard an advertisement on radio and TV from except the "free publicity" given to 'Teal' candidates.
* Note that in the seat of Hughes, independent candidate Georgia Steele, stood for the same issues of climate change and an integrity commission, but was not a member of the "Teal" group. She advertised on 2GB, but did not secure a seat.
Source: https://tallyroom.aec.gov.au/HouseDivisionPage-27966-124.htm
So the lessons are:
* Mustering active supporters for the candidates who door knock in the electorate. The more volunteers the easier it is to win. Old-fashioned door knocking may work to oust the incumbent member or by successive election campaigns until the candidate succeeds.
* Advertising alone does not work as demonstrated by the UAP and the indpendent candidate for Hughes.
* Workshopped media handling simple messages that can be handled within a seven second grab.
* Radio and TV advertising does not necessarily secure a seat. Mr Palmer's UAP spent $100 million and did not secure a seat. A candidate is more likely to succeed if resources were directed at the electorate level with active supporters rather than TV and radio advertisements.
* Targeted social media campaigns appear to work, especially with the Greens securing at least three lower house seats.
Media commentators have to do more in-depth analysis on why a commentators rather than parrotting public polls. It is easier to say "...the polls are pointing to a particular party winnning...". it is more interesting to know the whys of a candidate winning.
Thank you,
Anthony, look beyond the polls for a deeper analysis, Belfield in the land of the Wangal and Darug Peoples of the Eora Nation