Artificial Intelligence: What Consumers Say PDF

14d ago
0 Views
0 Downloads
271.36 KB
11 Pages
Transcription

Artificial Intelligence:what consumers sayFindings and policyrecommendationsof a multi-countrysurvey on AIThe Consumer Voice in Europe

IntroductionProducts and services consumers deal with on a daily basis – e.g.insurance policies based on automated risk assessments, product offers on e-commerce sites and price comparison tools– are increasingly powered by artificial intelligence (AI). Thistechnology promises to make consumers’ lives more convenient and our societies better. It can enable innovative productsand services, such as autonomous cars. It can also help makeprogress in areas such as scientific research and healthcare.But AI also has major implications on how consumers makechoices, how they interact with companies and the potentialrisks of products and services. AI and algorithm-based decisionmaking (ADM) are fundamentally altering the balance of powerbetween businesses and consumers. Consumers risk beingmanipulated and subjected to discriminatory treatment andarbitrary, non-transparent automated decisions. In a marketrun by algorithms, transparency and comparability for consumers may disappear.BEUC member organisations from across Europe conducteda survey to see what consumers think about AI. How much dothey know about it? Do they find AI useful? What are their mainconcerns? Is it properly regulated? What rights do they thinkthey have or should have?As the European Commission prepares new rules on AI, we hopethat the findings of this survey will inform EU decision-makersand help them to devise rules that guarantee strong consumerprotection that fully respect fundamental EU rights and values.02

Process andmethodologyThe survey was conducted simultaneously across nine EUcountries (Belgium, Denmark, France, Germany, Italy, Poland,Portugal, Spain and Sweden) throughout November andDecember 2019. The data, which was collected by a marketresearch company via an online questionnaire, was analysed bythe Belgian consumer organisation Test Achats/Test Aankoop.The questionnaire, comprising 22 questions, was given topanels of just over 1,000 respondents per country in Belgium,Italy, Spain and Portugal and panels of 1,500 in Denmark, France,Germany, Poland and Sweden. Respondents were selectedbased on age, gender and geographical region (according tothe distribution of the national general population). Sampleswere then weighed for age, gender, educational level andregion in order to be representative of the countries’ nationalpopulations. A postal questionnaire was also sent in Belgium,Italy, Portugal and Spain.For the purposes of this report we have structured the results ofthe survey around three topics: Consumers’ knowledge, perception and experience of AI Consumers’ concerns and trust in AI Consumers’ views on the regulatory framework for AI03

Main findingsConsumers consider AIcan be useful and thinkit will play an important rolein many areas of their lives1 A large majority of respondents perceive AI to be somewhat oreven very useful to them in different areas, especially when itis used to predict traffic accidents (91%), their health (87%) orfinancial problems (81%). A significant number of respondents (44% in Portugal and50% in Spain) also believe AI will contribute to increase our lifeexpectancy. Consumers also believe that AI will help to make the worldmore sustainable: this is the case for 44% of Spanish and 45%of Italian respondents.04

While they seebenefits, consumershave low trust in AI andits added value, as well asconcerns such as the abuseof personal data and the useof AI to manipulate theirdecisionsConsumers do notthink that the currentrules efficiently regulate andoversee AI. Consumers thinkthey should be properlyinformed about automateddecisions and have a rightto say “no’’ to them. For example, 78% of respondents consider that real-time carjourney suggestions offer some or even a lot of added value,yet almost 45% believe home virtual assistants offer no addedvalue at all. The same is true for personalised content/advertisement on e-commerce platforms: 44% feel it has no addedvalue. Significant numbers of people (50% in Sweden and 55% inPortugal) do not think that current rules effectively regulateAI*.2 Consumers actually reported having experienced “bad service” in several areas. 41% have this impression regardinginformation for loan proposals based on automated decisionsand 40% give the lowest score to the availability of options incustomer information/support systems run by ADM. There is a significant lack of trust: when asked whether theytrust if their privacy is protected when using AI devices, alarge majority of consumers state this to be medium or low.An average 50% of Belgian, Italian, Portuguese and Spanishas well as 45% of Danish, French, German, Polish and Swedishrespondents have low trust in the protection of their privacywith wearables*. With voice/virtual assistants, the level of lowtrust ranges from 40% in Sweden to 71% in Belgium.3 Around 56% of respondents in all countries – with peaks of70% in Belgium – have low trust in authorities to exert effectivecontrol over AI*. The majority of people (up to 83% in Spain) think that consumers should be well informed when they deal with an automatic decision system*. Respondents agree or strongly agree that AI users should havethe right to say “no” to automated decision-making. The numbers are as high as 78% in Italy and Portugal and 80% in Spain*. The majority of respondents (60%) in Belgium, Italy, Portugaland Spain also say that they agree or strongly agree that AI willlead to more abuse of personal data. For Denmark, France,Germany, Poland and Sweden it is 45%*. In Belgium, Italy, Portugal and Spain most people (64%) agreeor strongly agree that companies are using AI to manipulateconsumer decisions. The same is true for respondents fromFrance, Denmark, Germany, Poland and Sweden (52%)*.For some of the questions, respondents were asked to express whether they “strongly agreed’’, “agreed”, “disagreed” or “strongly disagreed” with certainstatements. For these questions, respondents in Denmark, France, Germany, Poland and Sweden also had the possibility to answer with “no opinion”. Thisoption was not available to respondents of the postal questionnaire in Belgium, Italy, Portugal and Spain. For this reason, results for the two groups of countriesare split in some sections of this report. This is indicated throughout the report with an asterisk (*).05

Detailed findings,assessment andrecommendationsHere we detail the findings from the survey, andgive our assessment and recommendations foraction.Knowledge, perception and experience of AITo get a better understanding of what AI means to consumers and their level of awareness, respondents were askedquestions about their knowledge, perception and experience of AI.Main findings 82% have previously heard of AI. Highlyeducated respondents are those whohave most heard of it (except for Spainwhere respondents’ financial situation isthe most important variable). While 48% say that they feel somewhatinformed about AI, only 9% say they arewell-informed. 43% say they are eitherill-informed about AI or have neverheard of it. 18% of people see AI as very presentin their daily lives while 39% see it assomewhat present. 22% think AI is verylittle or not at all present; 21% have neverheard of AI or have no idea about itspresence. Among respondents with no/06basic computer skills or no/basic skillswith connected devices, this increasesdrastically (up to 89%/84% in Swedenand 90%/90% in Denmark). Respondents are sceptical about AI’sability to perform highly complex humanbehaviour or human properties likeemotions. 41% and 45% of respondentsrespectively perceive that advancedtechnology will never be able to feelemotion or give a verdict in court. Yet55% and 69% respectively think thatadvanced technology is already ableto autonomously speak/interact andrecognise images and faces. Those whosay they are well informed about AIconsistently say that it is already beingused to autonomously perform imagerecognition as well as speaking/interacting (exception for Germany and Francefor image recognition). Respondents seem to hope that AI willhelp to resolve some fundamental problems of human life. In all participatingcountries, people find the followingservices based on machine calculationssomewhat or very useful:- Predicting traffic accidents (91%)- Predicting their health problems (87%)- Predicting their financial problems (81%)91.8% of all respondents find at least oneautomated service useful. I n all countries, those who haveadvanced automated devices skills ratethe usefulness of these services higher

than those with non/basic user skills.But, with one exception (Sweden, wherepredicting financial problems scoresslightly higher than predicting healthproblems for the non/basic users), alsonon/basic users rank the different services in the same order of usefulness.followed by image search engines (24%)and music identification apps 24%.However, in other areas – like home virtual assistants or personalised advertisements on shopping platforms – around45% say they have not experienced anyadded-value in the last 12 months. A considerable number of people recognise that AI services and applications,in certain areas, can have some addedvalue. But only car itinerary recommendations are rated by more than a quarterof respondents to have a lot of addedvalue (39%) among services peoplehave used in the past 12 months. This is C onsumers have not experiencedmany services/situations based onADM. Product purchase proposals rankhighest with only 18% (21% in Sweden;12% in France) and medical prescriptionslowest with 7% (12% in Poland; 5% inDenmark).Percentage of respondentswho find the followingservices based on machinecalculations somewhat/very useful. All DeFrnmiuarmk1009080706050403020100lgAlthough AI is still fairly new, consumersare already very much aware of it orsomewhat informed about it.Although almost all consumers surveyedare aware of AI, only half of them believeit is present in their lives. Not surprisingly,those who believe that it is most presentare also the most informed. In general,the direct connection between the levelof digital education and the appreciationof the benefits brought by AI is evident.Respondents are united in their belief thatautomation can be useful in the management of some activities and that they havealready perceived this usefulness at leastonce. The perception of this utility significantly changes according to the sectors.Despite this, only a few say that they haveactually benefited from it.Have you ever heard of AI ?BeBEUC assessmentand recommendationsNoYesThese results show that consumersbelieve that AI can bring benefits, butit is not delivering yet. They also showa need to further raise awareness andeducate consumers about what AI is,when and how it is used, what the risksare and what to do if something goeswrong. Consumer organisations canplay an important role in this by informing consumers about useful, safe, andlegally-compliant applications and howbest to use it.Did you experience any of theseservices based on AI ? Top salInsuranceproposalCustomerFinancialinformation / investmentsupportproposalNoYes07

Consumer concerns and trust in AIWe sought to understand and identify the main concerns consumers have about AI. This is essential to be able to addressthese concerns properly.Main findings* Regarding the use of AI by governments,57% of Belgian, Italian, Portuguese andSpanish respondents agree with thestatement that “governments are usingAI to control citizens”, with only 17%(strongly) disagreeing. The numberdrops to 44% for respondents fromDenmark, France, Germany, Poland andSweden. For Belgium, Italy, Portugal and Spainthe majority of respondents (60%) saythey agree or strongly agree that AI willlead to more abuse of personal data. ForDenmark, France, Germany, Poland andSweden it is 45%. Respondents have low trust in privacyprotection when using voice/virtualassistants. For instance, 71% of Belgianconsumers have low levels of trust, risingto 68% of German respondents. For Belgium, Italy, Portugal and Spain themajority of respondents (60%) agree or For Belgium, Italy, Portugal and Spain themajority of respondents (52%) say theystrongly agree or agree that AI is potentially dangerous because machines canfail. This is 46% in Denmark, France,Germany, Poland and Sweden.strongly agree that it is not clear who isaccountable if AI is not secure or causesharm. In Denmark, France Germany,Poland and Sweden this is 49%. The majority of respondents (51%) inBelgium, Italy, Portugal and Spain agreeor strongly agree that AI will lead tounfair discrimination based on individualcharacteristics or social categories. Thisis 37% in Denmark, France, Germany,Poland and Sweden.Respondents who (strongly) agreedthat companies use AI to manipulateconsumer decisionsHigh trustLow trustMedium trustendedlanSwanyPomGerHigh trustkDenmarceanFryanGermlandPoedSwlgBeMedium Low trust08cekDeRespondents in these countries weregiven the option to reply ‘No opinion’To what extent do you trust that your privacyis protected when using voice/virtual assistants y1009080706050403020100ium1009080706050403020100Fr In Belgium, Italy, Portugal and Spain themajority of respondents (64%) agree orstrongly agree that companies are usingAI to manipulate consumer decisions.The same is true for respondents fromDenmark, France, Germany, Poland andSweden (52%).No opinion

BEUC assessmentand recommendationsThe results of the survey show clearconcerns regarding the use of AI. Firstly,the majority of respondents consider AIwill increase the misuse of their personaldata. They particularly do not trust voice/virtual assistants to respect their privacy.Secondly, respondents also consider thatcompanies can use AI to manipulate theirdecisions and that AI will lead to unfairdiscrimination.Consumers have clear concerns regarding their privacy protection, AI manipulating their decisions, the risks ofdiscrimination, the reliability and safetyof AI, and the allocation of responsibilityand liability if something goes wrong.These concerns must be properlyaddressed to ensure strong protectionfor consumers and for them to trust thistechnology.Thirdly, respondents consider that AI canbe dangerous because machines can fail.Many are also worried that it is unclearwho is accountable if AI is not secure orcauses harm.The legal framework applicable to AI and the protectionof consumersThere are existing EU rules (i.e. data protection or consumer law) that apply to AI but no specific laws

Artificial Intelligence: what consumers say Findings and policy recommendations of a multi-country survey on AI. 02 Products and services consumers deal with on a daily basis – e.g. insurance policies based on automated risk assessments, pro - duct offers on e-commerce sites and price comparison tools – are increasingly powered by artificial intelligence (AI). This technology promises to ...