Content uploaded by Andreas Graefe
Author content
All content in this area was uploaded by Andreas Graefe on Oct 18, 2017
Content may be subject to copyright.
! 1!
!
Perception of Automated Computer-Generated News:
Credibility, Expertise, and Readability!
!
Paper!presented!at!the!11th!Dubrovnik!Media!Conference!Days:!!
Artificial!Intelligence,!Robots,!and!Media,!October!30–31!
!
Andreas!Graefea,*,!Mario!Haima,!Bastian!Haarmannb!!&!Hans-Bernd!Brosius!a!
!
a!Department!of!Communication!Studies!and!Media!Research,!LMU!Munich!
!
b!Fraunhofer!Institute!for!Communication,!Information!Processing!and!Ergonomics,!
Wachtberg,!Germany!
!
*!Send!correspondence!to!Andreas!Graefe,!Department!of!Communication!Studies!and!
Media!Research,!LMU!Munich,!Oettingenstraße!67,!80538!München,!Germany;!Phone:!
+49-89-2180-9466;!Email:!a.graefe@lmu.de!
!
Acknowledgments.! Matt! Carlson,! Christer! Clerwall,! Nick! Diakopoulos,! Arjen! van! Dalen,!
Konstantin!Dörr,!Bernhard!Goodwin,!Noam!Lemelshtrich!Latar,!Rasmus!Nielsen,!and!Neil!
Thurman!provided!helpful!suggestions.!We!also!received!suggestions!when!presenting!the!
manuscript!at!the!2015!Dubrovnik!Media!Days.!Dominik!Leiner!granted!access!to!the!SoSci!
Panel.! Veronika! Gburikova,! Steffi! Leupolt,! and! Dayana! Penkova! helped! conducting! the!
experiment.!
! 2!
!
Perception of Automated Computer-Generated News:
Credibility, Expertise, and Readability!
!
Abstract.! We! conducted! an! online! experiment! to! study! people’s! perception! of!
automated! computer-written! news.! Using! a! 2×2×2! design,! we! varied! the! article! topic!
(sports,! finance;! within-subjects)! and! both! the! articles’! actual! and! declared! source!
(human-written,!computer-written;!between!subjects).!986!subjects!rated! two!articles!on!
credibility,! readability,! and! journalistic! expertise.! Varying! the! declared! source! had! small!
but! consistent! effects:! subjects! rated! articles! declared! as! human-written! always! more!
favorably,! regardless! of! the! actual! source.! Varying! the! actual! source! had! larger! effects:!
subjects! rated! computer-written! articles! as! more! credible! and! higher! in! journalistic!
expertise!but!less!readable.!Across!topics,!subjects’!perceptions!did!not!differ.!The!results!
provide!conservative!estimates!for!the!favorability! of! computer-written! news,! which! will!
further! increase! over! time,! and! endorse! prior! calls! for! establishing! ethics! of! computer-
written!news.!
!
Keywords.!automated!journalism,!algorithmic!journalism,! journalism! ethics,! news!
perception,!robot!journalism,!Turing!test!!
!
!
! 3!
!
Introduction!
Although! computers! have! long! assisted! journalists! with! their! daily! work! (e.g.,! in!
researching!facts!and!analyzing!data),!journalists!have!remained!the!sole!creators!of!news.!
This!division!of!labor!is,!however,!currently!changing.!Advances!in!the!fields!of!information!
technology,! linguistics,! and! natural! language! generation! have! made! it! possible! for!
algorithms! to! autonomously! write! news! stories! from! data! that! is! stored! in! a! structured!
and!machine-readable!form.!This!development!is!sometimes!referred!to!as!automated!or!
robot! journalism! (Carlson,! 2015;! Clerwall,! 2014;! Lemelshtrich! Latar,! 2015;! Napoli,! 2014)!
and! is! part! of! a! larger! trend! known! as! computational! journalism,! which! describes! an!
increasing! influence! of! computation! and! data! on! journalism! (Anderson,! 2013;! Cohen,!
Hamilton,!&!Turner,!2011;!Lewis!&!Westlund,!2015).!
For!publishers,!the!most!obvious!benefit!of!computer-written!news!is!an!economic!
one:!computers!are!able!to!generate!news!at!a!much!larger!scale,!and!thus!at!a!lower!cost,!
than! human! journalists! (van! Dalen,! 2012).! Today,! companies! such! as! Narrative! Science!
and! Automated! Insights! provide! algorithms! that! generate! millions! of! articles! on! topics!
such!as!sports,!finance,!and!marketing!(Ulanoff,!2014),!and!publishers!have!already!begun!
to! use! computer-written! stories! in! their! news! coverage.! Forbes,! for! example,! has! been!
using!this!technology!since!2012!to!report!on!company!earnings!(Levy,!2012).!In!addition!
to!economic!benefits,!companies! are! seeking! to!take!advantage!of!the!speed!with!which!
computers!can!generate!news.!A!recent!example!is!the!Associated!Press,!which!published!
a!computer-written! report!on!Apple’s!quarterly!earnings!only!minutes!after!the!company!
released!its!figures!in!January!2015!(AP,!2015).!
! 4!
!
The! computer-written! news! industry! is! expected! to! grow! quickly.! Kristian!
Hammond,! co-founder! of! Narrative! Science,! predicted! that! computers! will! write! more!
than!90!percent!of!news!by!2025! (Levy,! 2012).! While! this! number! is!certainly!debatable,!
automated!journalism!is!likely!to!disrupt!news!writing!in!the!years!to!come.!The!expected!
growth! in! computer-written! content! will! result! in! additional! news! that! is! not! available!
today,!since!computers! will! report! on! small-scale! events!that!journalists! are! unwilling! to!
cover,!or!for! which! publishers! are!unwilling!to! hire! journalists.! Besides! simply!increasing!
the!quantity!of!news,!the!growth!of!computer-written!news!is!also!expected!to!affect!how!
journalists!create,!and!how!consumers!perceive,!news.!!
Scholars! have! only! just! begun! to! study! the! implications! of! this! development.! In!
particular,!researchers!have!looked!at!how!the!widespread!adoption!of!computer-written!
news! may! potentially! impact! the! traditional! role! of! journalists! and! the! quality! of! news!
coverage! in! general.! While! some! argue! from! a! theoretical! point! of! view! (Lemelshtrich!
Latar,!2015;!Napoli,!2014),!two!studies!analyzed!journalists’!news!coverage!of!Automated!
Insights!(van!Dalen,!2012)!and!Narrative!Science!(Carlson,!2015),!two!leading!companies!in!
automated!news!generation.!!
Regarding! the! potential! impact! on! journalists’! traditional! roles,! some! take! a!
positive! view! in! that! computer-written! news! will! aid! journalists! in! their! daily! work.! For!
example,! journalists! could! delegate! routine! tasks! (e.g.,! reporting! company! earnings! or!
recapping!sport!events)!to!algorithms.!Computer-written!stories! could!also!provide!a!first!
draft!that!covers!basic! facts,! which! journalists! would!then!verify!or! enrich! with! more!in-
depth!analyses!and!interpretation.!As!a!result,! so!the!theory!goes,!journalists!would!have!
more! time! available! for! higher! value! and! labor-intensive! tasks! such! as! investigative!
! 5!
!
reporting! (van! Dalen,! 2012).! An! example! is! crime! reporting! by! the! L.A.! Times! Homicide!
Report,!in!which!an!algorithm!provides! basic! facts,! such! as! the! date,! location,! time,!age,!
gender,! race,! and! jurisdiction! of! a! homicide.! In! a! second! step,! journalists! can! add! to! a!
story!by!providing!details!about!the!victim’s!life!and!family!(Young!&!Hermida,!2015).!!
Those! with! a! more! skeptical! view! suggest! that! the! increasing! adoption! of!
computer-written!news! will!put!pressure!on!journalists.!In!particular! those!who!currently!
perform! routine! tasks! in! areas! with! highly! structured! and! rich! databases! (e.g.,! sports,!
finance,! weather)! will! likely! be! unable! to! compete! with! automatic! data! collection! and!
writing!(Carlson,!2015).!!
Apart! from! publishers’! economic! considerations,! the! question! of! whether!
algorithms!will!augment!or!supplant!journalists! will!likely!depend!on!an!article’s!topic,!its!
purpose,! and! the! consumers’! expectations! regarding! the! quality! of! the! content.! That! is,!
while!algorithms!can!provide!answers!to!clearly!formulated!questions!by!analyzing!given!
data,! they! cannot! raise! questions! on! their! own! or! provide! opinions!on! important! social!
issues! or! proposed! policy! changes,! at! least! not! yet! (Lemelshtrich! Latar,! 2015).! Skeptics!
also! predict! that! news! consumers! would! dislike! reading! computer-written! stories.! The!
reason!is! that! algorithms! are! limited!in! understanding! and! producing! nuances!of! human!
natural!language!such!as!humor,!sarcasm,!and!metaphors.!As!a!result,!computer-written!
stories!tend!to!sound!technical!and!boring!(Lemelshtrich!Latar,!2015).!!
Proponents,! however,! argue! that! the! ability! of! algorithms! to! generate! natural!
human! language! will! improve,! which! will! make! the! content! more! appealing! to! news!
consumers.! More! importantly,! computer-written! news! could! potentially! increase! the!
quality!and!objectivity!of!news!coverage.!One!argument!is!that!computers!never!get!tired.!
! 6!
!
Thus,!algorithms!are!less!error-prone,! as! they! do! not! make!mistakes!such!as!overlooking!
facts.!Another! argument!is!that! algorithms!strictly!follow!pre-defined!rules!for!converting!
data!to!text!and!are!thus! capable!of!an!unbiased! account!of!facts.!The!latter!argument!is!
based! on! the! assumption! that! the! underlying! data! are! complete! and! correct! and,! more!
importantly,! the! algorithms! are! programmed! correctly! and! without! bias.! This! view,!
however,!is!rather!optimistic:!like!any!other!model,!algorithms!for!generating!computer-
written!news!rely!upon!data!and!assumptions,!which!both!are!subject!to!biases!and!errors!
(Lazer,! Kennedy,! King,! &! Vespignani,! 2014).! As! a! result,! algorithms! could! produce!
outcomes! that! were! unexpected! and! unintended! (Diakopoulos,! 2015).! First,! the!
underlying! data! may! be! wrong,! biased,! and! incomplete.! Second,! the! assumptions! built!
into!the!algorithms!may!be!wrong!or!could!reflect!the!conscious!or!unconscious!biases! of!
those!who!developed!or!commissioned!them!(Lemelshtrich!Latar,!2015).!!
Given! that! developers! are! unlikely! to! fully! disclose! their! algorithms,! it! remains!
unclear! how! to! evaluate! the! quality! of! the! algorithms! that! generate! computer-written!
articles.!A!promising!yet! complex! approach! might! be! reverse!engineering,!which!aims!at!
decoding!an!algorithm’s!set!of!rules!by!varying!certain!input!parameters!and!assessing!the!
effects! on! the! outcome! (Diakopoulos,! 2015).! An! alternative! approach! is! to! analyze! how!
news! consumers! perceive! the! quality! of! computer-written! news! in! relation! to! human-
written!news.!This!approach,!which!can! be!regarded!as!a!Turing!test!of!journalism,!is!the!
route!taken!in!the!present!study.!In!particular,! we! build! on—and! extend—prior! work! on!
how! recipients! perceive! computer-written! news.! In! the! remainder! of! this! paper,! we!
review!previous!studies!and!report!new!evidence!from!an!online!experiment!conducted!in!
Germany!with!986!participants.!!
! 7!
!
Prior!Evidence!
Prior!evidence!on!the!perception!of!computer-written!articles!is!limited.!We!are!aware!of!
only!two!studies,!which!differ!in!their!experimental!designs!(see!Table!1).!!
!
---!Table!1!about!here!---!
!
Clerwall! (2014)! analyzed! differences! in! perceived! quality,! measured! as! credibility!
and!readability,!of!news!articles.! He! presented! 46!Swedish!undergraduates!in!media!and!
communication! studies! with! one! of! two! versions! of! an! article! on! an! American! Football!
game.! The! article! was! generated! either! by! a! journalist! or! by! a! computer! but! the!
experiment!participants!were!not!given!any!information!about!the!source!(i.e.,!treatments!
5!and!6!in!Table!1).!This!setting!thus!reflected!a!situation!in!which!publishers!do!not!byline!
news! stories,! which! is! not! uncommon! for! wire! stories! (e.g.,! Associated! Press)! and!
computer-written!news!(Ulanoff,!2014).!The!articles!were!written!in!English,!contained!no!
pictures,! and! were! approximately! of! the! same! length.! After! reading! and! assessing! the!
article’s!credibility!and!readability,!participants!also! had!to!guess!whether!the!article!was!
written!by!a!journalist!or!generated!by!a!computer.!!
Overall,!differences!were! small,!which!is!not!surprising!given! the! sample!size,!and!
participants!were!unable!to!correctly!identify!the!article!source.!However,!the!direction!of!
effects!revealed!that!the!computer-written!articles!received!higher!ratings!on!credibility,!
whereas!the!articles!written!by!the!journalist!scored!higher!on!readability.!!
The! results! might! surprise.! Communication! students—who! one!could! expect! to!
have! a! higher! level! of! media! literacy! than! average! news! consumers—were! unable! to!
! 8!
!
distinguish!between! articles! generated! by! a! computer! and!those! written! by! a!journalist,!
and!even!favored!the!computer-written!article!in!terms!of!credibility.!!
Often,!however,!news!consumers!are!informed!whether!an!article!was!written!by!a!
journalist! or! created! by! a! computer.! Forbes,! for! example,! adds! the! byline! “by! Narrative!
Science”! to! computer-written! articles.! Similarly,! Associated! Press! reveals! if! a! story! was!
computer-written!(e.g.,!“This!story!was!generated!by!Automated!Insights!using!data!from!
Zacks!Investment!Research”).!It!remains!unclear,!though,!whether!consumers!understand!
the!meaning!of!such!bylines.!!
Thus,! one! might! ask! whether! perception! changes! if! consumers! know—or! think!
they!know—that!a!computer!generated!an!article.!This!question!was!addressed!by!van!der!
Kaa! and! Krahmer! (2014),! who! studied! the! perceived! credibility,! measured! as!
trustworthiness! and! journalistic! expertise,! of! computer-written! articles.! The! authors!
presented! 232! native! Dutch! speakers! (168! regular! news! consumers! and! 64! journalists)!
with!a!computer-written! article! that! either! reported!the!results! of! a! sports! event!(i.e.,!a!
soccer!game)!or! financial!news!(i.e.,!stock!prices).!The! articles!were!written!in!Dutch! and!
contained!no! pictures.!The!authors!then!manipulated!the!byline!of!the!article,!which!was!
either!correctly!declared!as!“written!by!a!computer”!or!wrongly!declared!as!“written!by!a!
journalist”!(i.e.,!cases! 3!and!4!in!Table! 1).!Among!regular!news!consumers,! differences!in!
perceived!expertise!and!trustworthiness!were!small,!although!articles!declared!as!written!
by!the!computer! received!slightly!higher!ratings! on!both!dimensions!than!those! declared!
as!written!by!the! journalist.! In!comparison,!the!64!journalist!participants!assigned! higher!
ratings! of! trustworthiness! to! articles! that! were! declared! as! written! by! the! journalist;! no!
differences!were! found! for! journalists’! perceptions!of! expertise.! The! study!also! revealed!
! 9!
!
some!differences!regarding!the!story!topic:!while!there!were!no!significant!differences!on!
expertise,!the!soccer!articles!scored!lower!on!trustworthiness!than!the!finance!articles.!!
In! sum,! the! results! conformed! to! those! of! Clerwall! (2014).! There! was! little!
difference!in!news! consumers’!perceived!credibility!of!articles,! regardless!of!whether!the!
articles! were! labeled! as! written! by! a! computer! or! by! a! journalist.! When! discussing!
potential! reasons! for! the! small! differences,! van! der! Kaa! and! Krahmer! (2014)! suggested!
that! news! consumers’! initial—perhaps! subconscious—expectations! might! influence! the!
results! in! favor! of! computer-written! articles.! In! particular,! subjects! might! have! high!
expectations! when! reading! an! article! (declared! as)! written! by! a! journalist,! and! low!
expectations! when! reading! an! article! (declared! as)! computer-written.! If! subjects! are!
positively! surprised! by! the! quality! of! a! computer-written! article! (and! several! of! their!
experiment! participants! reported! that! they! were),! they! might! assign! higher! ratings.! In!
contrast,!if!subjects’!expectations!in!the!quality!of!a!human-written!article!are!not!fulfilled,!
they!might!assign!lower!ratings.!!
The! present! study! addresses! this! question! by! building! on—and! extending—the!
experimental!design!of!the!two!previous!studies.!In!particular,!we!mimic!the!design!of!van!
der!Kaa!and!Krahmer!(2014)!by!varying!the!declared!article!source.!However,!we!also!vary!
the!actual! source!as!either!human-!or!computer-written.!This! variation!allows!us!to,!first,!
analyze! whether! news! consumers’! perceptions! are! indeed! influenced! by! their! initial!
expectations!regarding!the!(declared)!article!source.!If!so,!they!should!rate!human-written!
articles! higher! if! they! are! wrongly! declared! as! computer-written.! More! importantly,!
second,!this!experimental!design!enables!us!to!study!differences!in!perceptions!of!articles!
that! are! written! by! journalists! and! correctly! declared! as! such! (case! 1! in! Table! 1)! and!
!10!
!
articles!that! are!computer-written!and! correctly!declared!as! such!(case!4! in!Table!1).! This!
question,!which!has! not! been! analyzed! in! prior! research,!is!important!as! it! has! decision-
making!implications!for!publishers,!who!need!to!decide!whether!to!hire!human!journalists!
or!have!algorithms!produce!articles!on!certain!topics.!Finally,!our!setting!allows!for!testing!
whether!the!previous!findings!hold!in!a!different!country!(i.e.,!Germany).!
Method!
This! section! describes! our! measures! of! news! perception,! the! experimental! design,! the!
online! questionnaire,! the! participants,! and! the! stimulus! material.! For! additional!
information,!such!as! the! complete! questionnaire! and!the!full!article! texts! see! the! online!
appendix!at!http://bit.ly/18JGwsx.!
Measuring!news!quality!
Quality!of!news!is!a!fuzzy!concept!and!difficult!to!measure!as!it!means!different!things!to!
different!people.!For!example,!some!people!might!assess!an!article’s!quality!based!on!the!
excellence! of! writing! (e.g.,! the! use! of! stylistic! devices),! whereas! others! might! focus! on!
whether! the! article! is! well! researched! or! fits! their! view! of! the! world.! Thus,! perceived!
quality! does! not! necessarily! relate! to! objectively! measured! quality! (Urban! &! Schweiger,!
2014).! Nevertheless,! analyzing! news! consumers’! perceptions! of! different! aspects! of!
quality,!such!as!credibility,!is!common!in!studies!of!news!quality.!
Since! Hovland,! Janis,! and! Kelley! (1953),! numerous! scholars! have! evaluated! the!
quality!of!news.!Building!upon! seminal! work! by! Meyer! (1988)! and!West!(1994),!scholars!
have! distinguished! between! the! credibility! of! a! news! article’s! source! (e.g.,! Flanagin! &!
Metzger,! 2008;! Metzger,! Flanagin,! &! Medders,! 2010),! its! message! (e.g.,! Westerman,!
!11!
!
Spence,!&!Van!Der!Heide,!2014),!and!its!distributing!medium!(e.g.,!Golan,!2010;!Johnson!&!
Kaye,!2004).!Thereby,!researchers!commonly!use! factor! analysis! on! a! variety! of!items!to!
identify! different! dimensions! of! news! quality.! For! example,! Sundar! (1999)! presented!
people! with! news! articles! and! asked! for! their! open-ended! quality! evaluations,! which! he!
condensed!to!21!different!items.!He!then!validated!the!21!items!by!obtaining!ratings!for!
different! articles! from! different! subjects.! This! procedure! identified! four! central! factors!
that! people! consider! when! evaluating! news! content:! credibility,! readability,! quality,! and!
representativeness.!!
More! generally,! researchers! have! shown! that! news! quality! builds! upon! multiple!
dimensions! and! that! its! adequate! measure! requires! using! items! that! match! the! specific!
type!(e.g.,!written!or!audio!news)!and!topic!(e.g.,!sports!or!finance)!of!the!news!(Kohring!&!
Matthes,! 2007).! As! a! result,! the! literature! lacks! specific! guidelines! for! how! to! measure!
news!quality!and,!not! surprisingly! then,! different! researchers! have! used! different!scales.!
Looking!at!the!two!previous!studies!on!perception!of!computer-written!news!we!can!see!
such! differences! at! work.! Clerwall! (2014)! measured! quality! by! obtaining! readers’!
perceptions!of! credibility! (i.e.,!whether! they! found! the! articles! informative,! trustworthy,!
objective,!and!descriptive)!and!readability! (i.e.,! whether! they! found! the! articles! pleasant!
to!read,!clear,!well!written,!coherent,!and!not!boring).! Since! his! experimental! design! did!
not!reveal! the!article!source,! he!essentially!measured! message!credibility.!In! comparison,!
van!der!Kaa!and!Krahmer!(2014)!revealed!an!(either!true!or!false)!article! source!and!thus!
measured! message! and! source! credibility! by! obtaining! ratings! for! trustworthiness! (i.e.,!
reliability,! honesty,! accuracy,! and! fact-based)! and! journalistic! expertise! (i.e.,! expertise,!
intelligence,!authority).!!
!12!
!
In! order! to! develop! a! measure! of! content! perception! for! the! present! study,! we!
followed! an! approach! similar! to! Sundar! (1999).! We! conducted! a! pre-test! with! 40!
participants,! who! were! asked! to! rate! a! computer-written! article! (either! on! soccer! or!
finance)! on! 21! items,! using! a! 7-point! scale! from! “I! completely! agree”! to! “I! completely!
disagree.”!These!included!17!items!from!Sundar!(1999)!–!except!for!indifferent!items!and!
items! that! relate! to! “representativeness”,! as! this! dimension! does! not! cover! single! news!
items! but! rather! selections! of! news! (e.g.,! a! front! page)! –! as! well! as! four! items!
(trustworthy,! complete,! descriptive,! fact-based)! previously! used! by! van! der! Kaa! and!
Krahmer! (2014)! and! Clerwall! (2014).! In! addition,! respondents! could! suggest! new! items.!
However,!because!only!four!respondents! made! suggestions! and! there!was!no!coherence!
in! their! responses,! we! dismissed! the! open-ended! answers.! Exploratory! factor! analysis!
yielded!three!dimensions,!each! of! which!is!based!on!four!items!that! capture! perceptions!
of! credibility! (accurate,! trustworthy,! fair,! reliable),! readability! (entertaining,! interesting,!
vivid,! well-written),! and! journalistic! expertise! (coherent,! concise,! comprehensive,!
descriptive).!!
Stimulus!material!
In!a!similar!way!to!van!der!Kaa!and!Krahmer!(2014),!we!selected!articles!from!the!domains!
of! sports! and! finance,! which! are! representative! of! the! current! use! of! computer-written!
news.! The! reason! is! that,! for! both! topics,! reporting! is! commonly! fact-based! and! the!
underlying!data!(e.g.,!game!statistics,!historical!and!current! stock! prices,! etc.)! are! widely!
available!in!a!structured!format.!The!Fraunhofer!Institute!for!Communication,!Information!
Processing!and!Ergonomics!provided!us!with!the!computer-written!articles,!one!each!for!
sports!and! finance.! Their! software!is! currently! being! used! by!Finanzen100.de,! a! German!
!13!
!
website!that!publishes!financial!reports!and!which! is! part! of! the! Focus!Online!group.!For!
more! information! about! the! technology! see! Haarmann! and! Sikorski! (2015).! The! articles!
were!written!in!German!and!referred!to!events!from!the!previous!week.!In!order!to!avoid!
high!involvement,!the!soccer!article!reported!on!a!game!from!Germany’s!second!division.!
The!finance!article!reported!on!a!German!car!producer’s!share!performance.!!
The! human-written! stories! covered! the! same! events! and! were! obtained! from!
popular!German!websites!for!sports!(i.e.,!sport1.de)!and!finance!(i.e.,!deraktionaer.de,!the!
online! edition! of! a! weekly! financial! magazine).! To! assure! external! validity,! none! of! the!
articles! were! edited! or! shortened,! except! for! removing! pictures.! We! also! presented! the!
different!versions!to! nine! master! students! in!communication! science! and! asked! them!to!
identify!the!computer-written!articles;! the! students! were!correct!about!50%!of!the! time,!
which!reflects!random!guessing.!
Experimental!design!
The! experimental! 2×2×2! design! draws! on! ideas! from! both! previous! studies.! First,! in! a!
similar!way!to!Clerwall!(2014),!we!varied!the!actual!article!source!(i.e.,!whether!the!article!
was!actually!written!by! journalist! or!was!generated!by!a!computer).! Second,! like!van!der!
Kaa! and! Krahmer! (2014),! we! varied! the! declared! article! source! by! adding! a! byline! that!
labeled! the! article! as! computer-! or! human-written.! As! shown! in! Table!1,! this! between-
subjects!design!results!in!four!treatments:!!
!
(1)!Human-written!articles!correctly!declared!as!such;!!
(2)!Human-written!articles!wrongly!declared!as!computer-written;!!
(3)!Computer-written!articles!wrongly!declared!as!human-written;!!
!14!
!
(4)!Computer-written!articles!correctly!declared!as!such.!!
!
Each! participant! was! presented! one! article! of! each! topic,! in! randomized! order!
(within-subjects!design).!That!is,!participants!either!saw!a!soccer!article!first,!followed!by!a!
finance! article,! or! vice! versa.! Moreover,! each! participant! saw! one! article! declared! as!
written! by! an! algorithm! and! one! article! declared! as! written! by! a! journalist;! the! actual!
article! source! was! randomized.! In! other! words,! if! a! participant! were! assigned! to! either!
scenario!(1)!or!(3)!for! the! first! article,!then!she!was!assigned!to!either! scenario! (2)!or! (4)!
for!the!second!article,!and!vice!versa.!
Questionnaire!
In! an! online! questionnaire,! participants! were! first! asked! about! media! usage! patterns,!
journalistic! experience,! and! interest! in! different! topics.! Then,! participants! entered! the!
experimental!setting,!where!they!had!to!rate! both!articles!on!the! 12!measures!described!
above,!using!a!5-point!scale!ranging!from!“I!completely!agree”!to!“I!completely!disagree.”!
Finally,!participants!were!asked!for!socio-demographic!information.!Participants!spent,!on!
average,!8.5!minutes!(SD!=!3.2)!completing!the!questionnaire.!
Subjects!
Participants! were! recruited! through! the! SoSci! Panel,! a! noncommercial! online! access!
convenience!panel,!whose!approximately!90,000!active!members!voluntarily!participate!in!
scientific! surveys.! The! panel! has! two! major! advantages! compared! to! traditional! student!
samples.! First,! it! allows! for! the! recruitment! of! a! large! number! of! participants! and! thus!
addresses! the! small! sample! problem! of! previous! studies.! Second,! the! resulting! samples!
!15!
!
are! more! heterogeneous! than! student! samples! regarding! age,! geography,! professional!
background,!and! personal!interests.!However,!the!panel!members!are!not!representative!
of! the! German-speaking! population.! In! particular,! they! are! better! educated! than! the!
general!public.!In! addition,! given!the!exclusive!use!of!online! surveys,! the!panel!members!
have!a!generally! high! affinity!for!the!Internet,!which!makes! them! a!suitable!target!group!
for!the!present!study.!For!more!information!on!the!SoSci!Panel!see!Leiner!(2014).!!
A! total! of! 1107!subjects! participated! in! the! study! in! December! of! 2014.! After!
removing! incomplete! questionnaires,! 986!subjects! (53%! female)! remained.! The! average!
age! was! 38!years,! 55%! had! at! least! a! university! degree.! There! were! no! statistically!
significant! differences! across! the! experimental! groups! in! terms! of! age,! gender,! media!
usage!patterns,!prior!journalistic!experience,!and!interest!in!sports!and!finance.!
Results!
Cronbach’s!α!suggests!that!our!measures!of!the!three!dependent!constructs!were!reliable!
(credibility:!α!=!.83;!readability:!α!=!.85;!expertise:!α!=!.76).!Figure!1!shows!the!results!over!
both!topics.! The!columns!show!the!mean!ratings!per!construct!and!group.!The!error!bars!
depict! 95%! confidence! intervals! and! thus! indicate! statistical! significance;! Appendix! A!
shows! results! from! a! multivariate! analysis! of! variance.! All! data! and! calculations! are!
publicly! available! in! the! online! appendix! at! http://bit.ly/18JGwsx.! Upon! publication,! the!
replication!data!will!be!made!available!at!the!Harvard!Dataverse.!
Effect!of!the!topic!
Finance! articles! scored! between! 0.1! and! 0.5! points! (on! the! five-point! scale)! lower! than!
sports!articles!on!each!of!the!three!dimensions.!The!direction!of!the!effects,!however,!was!
!16!
!
identical! across! both! topics.! We! therefore! merged! the! data! in! order! to! simplify! the!
presentation!of!our!findings.!Results!per!topic!are!available!in!Appendix!B.!
!
---!Figure!1!about!here!---!
Effect!of!the!declared!source!
The!effect!of!the! declared! source! was! consistent! across! the! three!quality!measures!(i.e.,!
credibility,! expertise,! and! readability).! That! is,! regardless! of! the! actual! source,! articles!
were! always! rated! higher! if! declared! as! written! by! a! journalist.! In! all! but! one! case,!
however,! differences! were! rather! small! and,! thus,! not! statistically! significant! (i.e.,! the!
confidence!intervals!overlapped).!The!one!exception!were! the! computer-written! articles,!
which! were! rated! substantially! higher! in! terms! of! readability! if! they! were! (wrongly)!
declared!as!written!by!a!journalist.!
Effect!of!the!actual!source!
The! actual! source’s! effect!on! people’s! perception! differed! across! the! three! constructs.!
Regardless! of! the! declared! source,! the! computer-written! articles! were! rated! as! more!
credible! and! higher! in! terms! of! expertise! than! the! human-written! articles.! For! example,!
the!correctly!declared! computer-written! articles! received! a! mean!rating!of! 3.8! on! the! 5-
point!scale,!which!is!a!quarter!of!a!point!(or!7%)!higher!than!the!corresponding!ratings!for!
correctly!declared!human-written!articles.!
In!contrast,!the!results! for! readability!showed!the!opposite!effect:!human-written!
articles!scored!significantly!higher!than!those! written! by! the! computer.!Differences!were!
particularly! large! if! the! article! source! was! declared! correctly:! the! mean! rating! for! the!
!17!
!
human-written! article! (2.9)! was! 0.7! points! (or! 34%)! higher! than! the! rating! for! the!
computer-written!article!(2.2).!
Discussion!
Our! findings! corroborate! those! of! two! previous! studies! (Clerwall,! 2014;! van! der! Kaa! &!
Krahmer,! 2014),! which! used! different! experimental! designs! with! different! measures! of!
news! quality,! were! conducted! in! different! countries,! and! were! based! on! substantially!
smaller! samples! of! participants.! First,! computer-written! news! tends! to! be! rated! higher!
than! human-written! news! in! terms! of! credibility.! Second,! news! consumers! get! more!
pleasure! out! of! reading! human-written! as! opposed! to! computer-written! content.! Third,!
differences! in! terms! of! perceived! credibility! and! expertise! tend! to! be! small.! A! possible!
explanation! for! the! small! differences! is! that! algorithms! strictly! follow! standard!
conventions! of! news! writing! and,! as! a! result,! computer-written! stories! reflect! these!
conventions.!Given!that!a!major!portion!of!news!writing!is!a!simple!recitation!of!facts!and!
often! lacks! sophisticated! storytelling! and! narration,! it! is! not! surprising! that! recipients!
rated! both! article! sources! as! rather! credible! and! expert.! Interestingly,! however,! the!
recipients! did! not! like! reading! either! type! of! article! very! much.! Although! the! human-
written! ones! were! rated! as! clearly! more! readable! than! computer-written! stories,! their!
average!rating! was! still! below! the! mid-point! of!the! five-point! scale.! One!explanation! for!
low! readability! ratings! might! be! that! sports! and! finance! are! boring! subjects! for! many!
people.! Another! explanation! might! be! that! the! results! indicate! a! general! dissatisfaction!
with!news!writing!for!such!topics.!
!18!
!
For! many! topics,! and! in! particular! routine! tasks,! publishers!will! increasingly! have!
the!opportunity!to!rely!on!services!that!create!computer-written!news,!rather!than!hiring!
a!journalist!to!write!a!story.!Thus,!an!important!comparison,!which!has!not!been!analyzed!
in!prior!research,!is!to!compare!news!consumers’!perception!of!articles!that!are!written!by!
journalists! and! correctly! declared! as! such,! with! articles! that! are! computer-written! and!
correctly!declared!as!such.!Our!results!show!that!subjects!rate!computer-written!articles!
slightly! higher! in! terms! of! credibility! and! journalistic! expertise,! whereas! human-written!
articles! score! significantly! higher! in! terms! of! readability.! Given! the! current! state! of! the!
technology! and! apart! from! economic! considerations,! publishers! thus! face! a! trade-off!
between!credibility!and!readability!when!deciding!between!computer-!and!human-written!
stories.! That! said,! the! readability! of! computer-written! news! is! likely! to! further! improve!
over! time,! as! computer! linguists! are! constantly! enhancing! their! algorithms’! ability! to!
analyze!large!datasets!and!to!generate!natural! human!language!such!as! humor!or!poetry!
(Gonçalo!Oliveira!&!Cardoso,!2015;!Petrovic!&!Matthews,!2013).!In!comparison,!it!is!rather!
unlikely! that! the! quality! of! human! journalists! will! equally! improve—at! least! not! at! the!
same! pace.! In! the! short! term,! we! would! thus! expect! follow-up! studies! to! find! even!
stronger! effects! in! favor! of! computer-written! content.! However,! such! effects! may! not!
necessarily! persist! in! the! long-term.! It! may! well! be! that! after! readers’! initial! excitement!
with!the!new! technology,! algorithmic!news!that!builds!on!a! static! set!of!rules!might!”get!
old,”!in! particular,!if!used! at!a!large!scale.!If!so,!readers!may!be!drawn!towards!fresh!and!
creative!human!writing!styles!again,!which!may! create! new! opportunities! for! journalists.!
Future!research!should!track!how!the!quality!of!both!human-!and!computer-written!news!
!19!
!
will!evolve!and! how!people’s!expectations!towards!and!perceptions! of!such!content!may!
change!over!time.!
The!results!further! show! that! articles!are!consistently!perceived!more!favorably! if!
they!are!declared!as! written! by! a! human! journalist,! regardless!of!the!actual! source.! This!
finding! has! two! important! implications.! First,! the! results! address! the! question! raised! by!
van! der! Kaa! and! Krahmer! (2014),! who! suggested! that! consumers’! initial! expectations!
regarding! the! quality! of! the! declared! article! source! might! influence! their! perceptions! of!
quality.! In! particular,! they! argued! that! consumers! might! have! low! expectations! for!
computer-written!articles!and!might!thus!be!positively!surprised!by!their!quality,!which,!in!
turn,!would!lead!to!higher!ratings.!If!this!is!true,!then!human-written!articles!should!score!
higher! when! they! are! declared! as! computer-written.! Our! results! suggest! that! this! is! not!
the!case.!In!fact,!the!effects!pointed!in!the!opposite!direction:!human-written!articles!that!
were! wrongly! declared! as! computer-written! were! perceived! as! less! favourable! than! the!
same!article!correctly!declared!as!written! by!a!journalist.!Second,! although!differences!in!
effect! sizes! were! small,! the! results! might! tempt! publishers! to! assign! human! names! to!
computer-written!articles.!The!results!therefore!endorse!prior!calls!for!establishing!ethics!
of! robot! journalism! (e.g.,! Diakopoulos,! 2015;! Lemelshtrich! Latar,! 2015).! Publishers,! for!
example,!should!commit!to!faithfully!revealing!who!created!an!article.!!
In!sum,!the!available!evidence!suggests!that!the!quality!of!computer-written!news!
is!competitive! with! that! of! human! journalists! for!routine! tasks! for! which!there! are! well-
structured,! machine-readable,! and! reliable! data.! In! such! situations,! news-generating!
algorithms! excel! by! quickly! extracting! information! from! data,! weighting! information! by!
importance,!generating!news!narratives,!and!varying!writing!styles.!Popular!applications!of!
!20!
!
computer-written! news! currently! include! data-heavy! domains! such! as! weather!
forecasting,! sports! news,! traffic! reporting,! financial! analysis,! earthquake! warnings,! and!
crime!reporting!(Young!&!Hermida,!2015).!For!such!routine!tasks,!journalists!may!face!the!
danger!of!eventually!being!replaced!by!automated!journalism.!
It!is!important!to!note!that!our!results!cannot!be!generalized!to!topics!that!are!not!
solely! fact-based! and! for! which! journalists! contribute! value! by! providing! interpretation,!
reasoning,! and! opinion,! for! example,! when! it! comes! to! discussing! social! and! political!
issues.! Currently,! computer-written! stories! for! such! complex! problems! are! not! yet!
available.!However,!we!expect!that!the!quality!of!computer-written!news!will!continue!to!
improve,!which!might!enable!algorithms!to!generate!journalistic!output!other!than!simply!
reciting! facts.! Already! today,! algorithms! rely! on! pre-defined! rules! to! obtain! additional!
insights! from! the! data.! For! example,! for! previews! of! soccer! games,! the! algorithms! take!
into!account!the! teams’! historical!record!(against!each!other),!or! which!players!will!need!
to!avoid!a!booking!in!order!to!not!be!suspended!the!following!game.!!
Automated!journalism!may!comprise!a! major! share! of! news! writing!in!the!future.!
Due! to! the! ever-increasing! availability! of! data,! algorithms! will! be! able! to! cover! events!
where!currently!limited!news!is!available!(Carlson,!2015).!In!addition,!algorithms!will!likely!
be! able! to! write! stories! that! suit! individual! readers’! interests,! political! views,! and!
education!levels.!This!development!raises!questions!about!possible!implications!about!the!
future! of! journalism! and! its! relationship! to! the! democratic! process.! Given! the! idea! that!
journalism! (among! others)! should! enable! citizens! to! act! on! politically! well-informed!
grounds,!automated!journalism!could!both!hinder!and!foster!these!ideals.!On!the!positive!
side,!the! ability! to! personalize!news!may!make! it! possible! to!attract! a! broader! audience!
!21!
!
and!thus!increase! the!number!of!politically!informed! people.! Furthermore,!if!automation!
of! routine! tasks! will! indeed! free! up! resources,! journalists! might! have! more! time! for!in-
depth! analysis,! which! could! improve!news! quality.! On! the! negative! side,! an! increasing!
quantity!of!available!news!will!further!increase!people’s!burden!to!find!news!that!is!most!
relevant! to! them.! Given! news! consumers’! limited! time,! available! offers! need! to! be!
customized!to!individual! needs,!for!instance!by! increasingly!relying!on!personalized!news!
aggregators! such! as! Google! News! or! relevance-sorting! algorithms! such! as! Facebook’s!
news! stream.! Such! increasing! personalization! could! lead! to! fragmentation! or! “filter!
bubble”! (Pariser,! 2011)! effects! within! society.! The! concern! is! that! personalization! leads!
individuals!to!consume!more! and! more! of! the! same! information.!As!a!result,!people!are!
unlikely!to!consume!information!that!challenges!their!views!or!contradicts!their!interests,!
which!may!carry!risks!for!the!formation!of!public!opinion!in!a!democratic!society.!!
An!increasing!use!of!computer-written!news!also!raises!more!general!questions!as!
to! whether! we! can! and! should! trust! algorithms! as! a! mechanism! to! provide! checks! and!
balances,! to! identify! important! issues,! and! to! establishing! a! common! agenda! for! the!
democratic! process! of! public! opinion! formation.! For! instance,! algorithms! could! analyze!
publicly! available! information! such! as! annual! reports! or! speech! protocols! in! order! to!
provide! insights! on! political! and! economic! questions! faster! and! on! a! much! larger! scale!
than! any! human! journalist.! Thus,! automation! could!improve! transparency! and! point!
journalists!to!important!issues!that!need!further!attention.!On!the!other!hand,!algorithms!
for!generating!automated!news!rely! on! data! and! pre-defined! rules,! which! are! subject!to!
biases! and! errors.! Publishers! thus! have! to! assure! a! certain! level! of! algorithmic!
!22!
!
transparency! and! accountability! (Diakopoulos,! 2015),! in! particular,! if! the! technology! is!
used!for!critical!and!controversial!problems.!
!
!!
!23!
!
REFERENCES!
!
Anderson,! C.! (2013).! Towards! a! sociology! of! computational! and! algorithmic!
journalism.! New$ Media$ &$ Society,$ 15(7),! 1005-1021.!
doi:10.1177/1461444812465137!
AP.!(2015,!January!27).!Apple!tops!Street!1Q!forecasts.$Yahoo$Finance.!Retrieved!from!
http://finance.yahoo.com/news/apple-tops-street-1q-forecasts-213944804.html!
Carlson,! M.! (2015).! The! Robotic! Reporter.! Digital$ Journalism,$ 3(3),! 416-431.!
doi:10.1080/21670811.2014.976412!
Clerwall,! C.! (2014).! Enter! the! Robot! Journalist.! Journalism$ Practice,$ 8(5),! 519-531.!
doi:10.1080/17512786.2014.883116!
Cohen,! S.,! Hamilton,! J.! T.,! &! Turner,! F.! (2011).! Computational! journalism.! Commun.$
ACM,$54(10),!66-71.!doi:10.1145/2001269.2001288!
Diakopoulos,!N.!(2015).!Algorithmic!Accountability.!Digital$Journalism,$3(3),!398-415.!
doi:10.1080/21670811.2014.976411!
Flanagin,! A.! J.,! &! Metzger,! M.! J.! (2008).! Digital! media! and! youth:! Unparalleled!
opportunity!and!unprecedented!responsibility.!In!M.!J.!Metzger!&!A.!J.!Flanagin!
(Eds.),! Digital$ Media,$ Youth,$ and$ Credibility!(pp.! 5-27).! Cambridge,! MA:! MIT!
Press.!
Golan,! G.! J.! (2010).! New! perspectives! on! media! credibility! research.! American$
Behavioral$Scientist,$54(3),!3-7.!doi:10.1177/0002764210376307!
Gonçalo!Oliveira,!H.,! &!Cardoso,!A.!(2015).!Poetry! generation!with!PoeTryMe.!In!T.! R.!
Besold,!M.!Schorlemmer,!&! A.! Smaill!(Eds.),!Computational$Creativity$Research:$
Towards$Creative$Machines!(Vol.!7,!pp.!243-266):!Atlantis!Press.!
Haarmann,!B.,! &! Sikorski,!L.! (2015).! Natural$language$news$generation$from$big$data.!
Paper!presented!at! the!International!Conference!on! Knowledge!Discovery!and!
Information!Retrieval,!Copenhagen.!
Hovland,! C.! I.,! Janis,! I.! L.,! &! Kelley,! H.! H.! (1953).! Communication$ and$ Persuasion.$
Psychological$Studies$of$Opinion$Change.!New!Haven,!CT:!Yale!University!Press.!
Johnson,!T.!J.,!&!Kaye,!B.!K.! (2004).!Wag!the!Blog:!How!Reliance!on!Traditional!Media!
and! the! Internet! Influence! Credibility! Perceptions! of! Weblogs! Among! Blog!
Users.! Journalism$ &$ Mass$ Communication$ Quarterly,$ 81(3),! 622-642.!
doi:10.1177/107769900408100310!
!24!
!
Kohring,!M.,!&!Matthes,!J.!(2007).!Trust!in!news!media:!Development!and!validation!of!
a! multidimensional! scale.! Communication$ Research,$ 34(2),! 231-252.!
doi:10.1177/0093650206298071!
Lazer,! D.,! Kennedy,! R.,! King,! G.,! &! Vespignani,! A.! (2014).! The! parable! of! Google! Flu:!
Traps! in! big! data! analysis.! Science,$ 343(6176),! 1203-1205.!
doi:10.1126/science.1248506!
Leiner,!D.!(2014).!Convenience!samples!from!online!respondent!pools:!A!case!study!of!
the! SoSci! Panel.! Working$ paper,! Available! at!
https://www.researchgate.net/publication/259669050.!!
Lemelshtrich! Latar,! N.! (2015).! The! robot! journalist! in! the! age! of! social! physics:! The!
end! of! human! journalism?! In! G.! Einav! (Ed.),! The$ New$ World$ of$ Transitioned$
Media!(pp.!65-80):!Springer!International!Publishing.!
Levy,!S.!(2012).!Can!an!Algorithm!Write!a!Better!News!Story!Than!a!Human!Reporter?!
Wired.!!Retrieved! from! http://www.wired.com/2012/04/can-an-algorithm-write-
a-better-news-story-than-a-human-reporter/!
Lewis,!S.!C.,! &!Westlund,!O.!(2015).!Actors,! actants,!audiences,!and!activities!in! cross-
media! news! work.! Digital$ Journalism,$ 3(1),! 19-37.!
doi:10.1080/21670811.2014.927986!
Metzger,!M.!J.,!Flanagin,!A.!J.,!&!Medders,!R.!B.!(2010).!Social!and!Heuristic!Approaches!
to! Credibility! Evaluation! Online.! Journal$ of$ Communication,$ 60(3),! 413-439.!
doi:10.1111/j.1460-2466.2010.01488.x!
Meyer,! P.! (1988).! Defining! and! Measuring! Credibility! of! Newspapers:! Developing! an!
Index.! Journalism$ &$ Mass$ Communication$ Quarterly,$ 65(3),! 567-574.!
doi:10.1177/107769908806500301!
Napoli,! P.! M.! (2014).! Automated! Media:! An! Institutional! Theory! Perspective! on!
Algorithmic! Media! Production! and! Consumption.! Communication$ Theory,$
24(3),!340-360.!doi:10.1111/comt.12039!
Pariser,! E.! (2011).! The$Filter$Bubble:$What$the$Internet$is$Hiding$from$You.! New! York:!
Penguin!Press.!
Petrovic,!S.,!&!Matthews,!D.! (2013).!Unsupervised$joke$generation$from$big$data.!Paper!
presented! at! the! 51st! Annual! Meeting! of! the! Association! of! Computer!
Linguistics,!Sofia,!Bulgaria.!
Sundar,! S.! S.! (1999).! Exploring! receivers'! criteria! for! perception! of! print! and! online!
news.! Journalism$ &$ Mass$ Communication$ Quarterly,$ 76(2),! 373-386.!
doi:10.1177/107769909907600213!
!25!
!
Ulanoff,! L.! (2014).! Need! to! write! 5! million! stories! a! week?! Robot! reporters! to! the!
rescue.! Mashable.!!Retrieved! from! http://mashable.com/2014/07/01/robot-
reporters-add-data-to-the-five-ws/!
Urban,! J.,! &! Schweiger,! W.! (2014).! News! Quality! from! the! Recipients'! Perspective.!
Journalism$Studies,$15(6),!821-840.!doi:10.1080/1461670X.2013.856670!
van!Dalen,!A.!(2012).!The!algorithms!behind!the!headlines.!Journalism$Practice,$6(5-6),!
648-658.!doi:10.1080/17512786.2012.667268!
van!der!Kaa,!H.,!&!Krahmer,!E.!(2014).! Journalist$versus$news$consumer:$The$perceived$
credibility$ of$ machine$ written$ news.!Paper! presented! at! the!
Computation+Journalism!Symposium,!Columbia!University,!New!York!City.!
West,!M.!D.!(1994).!Validating!a!scale!for!the!measurement!of!credibility:!A!covariance!
structure!modeling!approach.!Journalism$Quarterly,$71(1),!159-168.!!
Westerman,!D.,!Spence,!P.!R.,!&!Van!Der!Heide,!B.!(2014).!Social!Media!as!Information!
Source:! Recency! of! Updates! and! Credibility! of! Information.! Journal$ of$
Computer-Mediated$Communication,$19(2),!171-183.!doi:10.1111/jcc4.12041!
Young,!M.!L.,!&!Hermida,!A.!(2015).!From!Mr.!and!Mrs.!Outlier!To!Central!Tendencies.!
Digital$Journalism,$3(3),!381-397.!doi:10.1080/21670811.2014.976409!
!
!26!
!
Figure 1: Mean ratings per group
Notes:
• Values represent means
• Error bars indicate 95% confidence intervals
• Treatments (1) to (4) as shown in Table 1
!
!
!!
!27!
!
Table!1:!Experimental!studies!of!perception!of!computer-generated!news!available!to!
date!
Article source declared as written/generated by
Journalist
Algorithm
Not declared
Actual article source
Human
(1)
(2)
(5)
(Clerwall, 2014)
Computer
(3)
(van der Kaa &
Krahmer, 2014)
(4)
(van der Kaa &
Krahmer, 2014)
(6)
(Clerwall, 2014)
Notes: treatments analyzed in the present study: (1), (2), (3), and (4)
!
!!
!28!
!
Appendix!A:!Main!effects’!analyses!of!variance!
Independent variable
Dependent variable
F
Pillai’s V
Effect size r
Actual Source
F(3, 1848) = 119.2 ***
.16
Credibility
F(1, 1850) = 75.8 ***
.20
Readability
F(1, 1850) = 146.1 ***
.27
Expertise
F(1, 1850) = 19.4 ***
.10
Declared Source
F(3, 1848) = 10.8 ***
.02
Credibility
F(1, 1850) = 6.5 *
.05
Readability
F(1, 1850) = 25.7 ***
.11
Expertise
F(1, 1850) = 1.9
-
Real × Declared
F(3, 1848) = 0.9
-
Credibility
F(1, 1850) = 1.0
-
Readability
F(1, 1850) = 0.8
-
Expertise
F(1, 1850) = 0.1
-
Notes: Results from one multivariate (overall) and three univariate (one per independent variable) analyses
of variance (AnOVa). Rows containing independent variables and Pillai’s V show multivariate results; rows
containing dependent variables show univariate analyses. Results depict three levels of statistical
significance (< .05 *; < .01 **; < .001 ***) and include all ratings (i.e., both topics; n = 1854).
! !
!29!
!
Appendix!B:!Results!per!topic!
N
Credibility
Readability
Expertise
Finance
921
Human-written
declared as journalist
232
3.3 (0.06)
2.8 (0.06)
3.0 (0.06)
declared as algorithm
228
3.1 (0.06)
2.6 (0.06)
3.1 (0.06)
Computer-generated
declared as journalist
231
3.7 (0.05)
2.5 (0.06)
3.2 (0.05)
declared as algorithm
230
3.7 (0.06)
2.3 (0.06)
3.1 (0.06)
Soccer
933
Human-written
declared as journalist
220
3.8 (0.05)
3.0 (0.07)
3.4 (0.05)
declared as algorithm
235
3.6 (0.05)
2.8 (0.06)
3.2 (0.05)
Computer-generated
declared as journalist
239
3.9 (0.05)
2.4 (0.06)
3.6 (0.05)
declared as algorithm
239
3.9 (0.05)
2.1 (0.05)
3.6 (0.05)
Mean ratings based on 5-point scales (1 = “I completely disagree”, 5 = “I completely agree”), standard errors in parentheses.
!