Chap e 2
Gene a i e AI as a acili a o o
delibe a e p ac ice in ansla o aining
E ik Angelone
TH Köln Uni e si y o Applied Sciences, Ge many
Ad ancemen along an expe ise ajec o y, whe he in aining o p o essional
con ex s, s ems om ansla o s ha ing ample oppo uni y o engage in delibe a e
p ac ice, de ined in he expe ise s udies li e a u e as s a egically designed ac i i-
ies o imp o e pe o mance (E icsson e al. 1993). In o de o p ac ice o be delibe -
a e, a numbe o co e condi ions need o be me , including, among o he s, in insic
lea ne mo i a ion, ocused and sel -di ec ed pe o mance moni o ing, in o ma-
i e (and ela i ely immedia e) eedback, asks being s uc u ed and unde aken
a an app op ia e di icul y le el, and lea ne s ha ing oppo uni ies o co ec e -
o s (Sh e e 2006). In ansla o aining con ex s, complex, he e ogeneous lea ne
p o iles, pa icula ly when i comes o knowledge, skills, and compe ences (EMT
2022), pose inhe en challenges in a emp s o make su e hese condi ions a e ad-
equa ely me . In add essing hese challenges, and in op imising oppo uni ies o
delibe a e p ac ice in gene al, Gene a i e AI o e s pedagogical alue as a ehicle
o guide sel -di ec ed lea ning. This chap e will discuss how gene a i e AI can be
used o acili a e delibe a e p ac ice in ansla o aining, whe e s uden s engage
in sel -di ec ed me acogni i e ac i i y o e lec on and assess ace s o hei own
pe o mance, and, in doing so, acqui e and ad ance hei expe ise.
1 In oduc ion
Technological ad ancemen , mos ecen ly in he ealm o gene a i e AI, is
apidly changing he oles and esponsibili ies o p o essional ansla o s. I is
also e-shaping he compe ences hey need o possess o ind success in he
language indus y, as e lec ed in he ecen ly upda ed EMT T ansla ion Com-
pe ence F amewo k (EMT 2022). In imes o seemingly pe pe ual change and a
E ik Angelone. 2026. Gene a i e AI as a acili a o o delibe a e p ac ice in ansla o
aining. In JC Pene , Joss Moo kens & Masa u Yamada (eds.), Teaching ansla ion
in he age o gene a i e AI: New pa adigm, new lea ning?, 27–47. Be lin: Language
Science P ess. DOI: 10.5281/zenodo.17641066
E ik Angelone
co esponding need o adap (Angelone 2023), one aspec ega ding compe ence
has emained ai ly cons an , namely he need o ansla o s o c i ically e lec
on, assess, and op imise hei own pe o mance. This capaci y is a icula ed in
se e al o he sub-compe ences pe aining o o e a ching ‘ ansla ion compe-
ence’ in he EMT amewo k, such as ansla o s being able o “analyse a sou ce
documen , iden i y po en ial ex ual and cogni i e di icul ies and assess he
s a egies and esou ces needed o e o mula e i ”, and “analyse and jus i y hei
ansla ion solu ions and choices” (EMT 2022: 8). I is also mani es in he domain
o pe sonal and in e pe sonal compe ence as he abili y o “manage wo kload,
cogni i e load, s ess and c i ical p o essional si ua ions”, and “con inuously sel -
e alua e, upda e and de elop compe ences and skills h ough pe sonal s a egies
and collabo a i e lea ning” (EMT 2022: 10).
Compe ence, which can b oadly be de ined as he se o knowledge and skills
needed o success ully ansla e, is closely ied o expe ise, which, depending on
esea ch pa adigm and heo e ical ounda ion, can be desc ibed as ‘consis en ly
supe io pe o mance’ in a gi en ask domain (E icsson & Cha ness 1994), o
‘op imal pe o mance’ ac oss a ied, bu in e ela ed ask domains (Ha ano & In-
agaki 1986). Pe o mance, when i comes o success ul ansla ion, in ol es no
only quali y (such as iden i ying and mi iga ing e o s o success ully adhe ing
o es ablished p ojec speci ica ions), bu also p oduc i i y (such as le e aging
assis i e echnologies o expedi e ou pu o u ilising e gonomically sound ap-
p oaches when acing ime cons ain s). Expe ise should be ega ded as some-
hing de elopmen al and inc emen al (Sh e e 2018) a he han a desi able ‘end
s a e’ in he sense o an ‘expe ’ ansla o . Wi h his in mind, aining o ans-
la ion expe ise acquisi ion and ad ancemen has an impo an place om he
ou se .
In o de o ad ance along an expe ise ajec o y, o example om ‘no ice’, o
‘ad anced beginne ’, o ‘compe en ’, o ‘p o icien ’, o ‘expe ’ (D ey us & D ey-
us 1986), expe ience alone will likely no su ice. Ins ead, ansla o s should in-
en ionally seek oppo uni ies o engage in ‘delibe a e p ac ice’, which can be
de ined as “indi idualized aining ac i i ies especially designed by a coach o
eache o imp o e speci ic aspec s o an indi idual’s pe o mance h ough epe-
i ion and successi e e inemen ” (E icsson & Lehmann 1996: 278–279). Delibe -
a e p ac ice necessi a es mo ing beyond “pla eaus whe e one is com o able and
con iden ” (Ho n & Masunaga 2006: 601). Unlike he somewha p edic able ac i -
i y and app oaches ha ansla o s migh be inclined o encoun e in hei daily
wo k, delibe a e p ac ice in ol es “explo ing al e na i e me hods wi h unknown
eliabili y” (E icsson e al. 1993: 368). Indeed, common indus y cons ain s on
ime, budge , in as uc u e, and isk- aking in gene al make delibe a e p ac ice
28
2 Gene a i e AI as a acili a o o delibe a e p ac ice in ansla o aining
some hing dis inc om ‘wo k’. I is mo e o an ex e nal, up on in es men ha
can ul ima ely op imise wo k pe o mance.
2 Co e condi ions o delibe a e p ac ice
In o de o p ac ice o be delibe a e, a numbe o impo an co e condi ions need
o be me . T ansla ion S udies schola s ha e ou lined hese as hey pe ain o he
ask o ansla ion (Sh e e 2006). Pe haps mos impo an is he condi ion o
ansla o s ecei ing immedia e, in o ma i e eedback on hei pe o mance. “In
he absence o adequa e eedback, e icien lea ning is impossible and imp o e-
men minimal” (E icsson e al. 1993: 367). In o ma i e eedback ideally encom-
passes aspec s o ansla ion ha a e ans e able and applicable ac oss asks.
Fo example, eedback on he e icacy o ine icacy o a gi en ex e nal esou ce
o assis i e echnology is likely mo e in o ma i e han eedback on a misspelled
wo d ha he ansla o migh encoun e in he con ex o one gi en ansla ion
bu ne e again.
As a second condi ion, ansla o s need o ha e oppo uni ies o co ec e -
o s. In pedagogical con ex s, his condi ion is o en me h ough wo kshopping,
whe e s uden s submi and ecei e eedback on d a s, ollowed by he oppo -
uni y o submi e isions in which e o s b ough o hei a en ion can be
co ec ed. Lea ning is pa icula ly enhanced when e o co ec ion s ems om
disco e y-based lea ning, whe e e o s a e anno a ed, bu delibe a ely no ully
spelled ou by aine s. The e y p esence o a aine is ye ano he co e con-
di ion o delibe a e p ac ice. The aine , o ‘coach’, is esponsible no only o
p o iding in o ma i e eedback, bu also o es ablishing indi idualised lea ning
objec i es (Mille e al. 2020) and designing asks based on i m unde s anding
o he lea ne ’s p e-exis ing knowledge. Some p oponen s o delibe a e p ac ice-
o ien ed aining sugges ha indi idualised supe ision is ul ima ely p e e ed
o e g oup-based ins uc ion (E icsson e al. 1993: 367).
The aine also plays a pi o al ole in ul illing ano he co e dimension o
delibe a e p ac ice, namely making su e he ansla o is engaging in asks o
an app op ia e di icul y le el. S agna ion, due o s ic adhe ence o he ied
and ue, can esul in a pla eau e ec ha s ands in he way o expe ise ad-
ancemen . T ansla o s may encoun e his when wo king on asks o a uni o m
di icul y le el (same sou ce ex leng h, same eadabili y le el, same se o ex-
e nal esou ces, same p ojec speci ica ions, same assessmen pa ame e s, e c.).
T aine s can push ainees o wo k ou side o hei com o zones and help de e -
mine “when ansi ions o mo e complex and challenging asks a e app op ia e”
(E icsson e al. 1993: 367).
29
E ik Angelone
Delibe a e p ac ice hinges on he lea ne ’s in insic mo i a ion, ano he co e
condi ion. T ansla o s looking o ad ance along an expe ise ajec o y need o
emb ace wo king ou side o hei com o zones and ha e a i m belie ha de-
libe a e p ac ice will ul ima ely imp o e hei pe o mance. Delibe a e p ac ice
akes immense ime and e o . The o -deba ed 10,000-hou ‘ ule’ o becoming
an ‘expe ’ (Gladwell 2011) is o en ci ed in a emp s o quan i y jus how much
ime and e o a e equi ed. This would amoun o “ wo and a hal yea s o sus-
ained e o ul p ac ice e e y day o i e hou s” (Sh e e 2019: 173–174), aising
ques ions ega ding easibili y o his ‘ ule’ and i s place in he delibe a e p ac ice
model.
Delibe a e p ac ice calls o he lea ne ’s commi men o “conscious pe o -
mance moni o ing” (Ho n & Masunaga 2006: 601) and engaging in ull concen-
a ion, as opposed o “mindless, ou ine pe o mance” (E icsson 2006: 692). This
akes sel -discipline and lea ne dedica ion o honing me acogni i e capaci ies.
S a egic sca olding by he aine , along wi h aine / ainee dedica ion o he
cen al ideas o cogni i e cons uc i ism (Piage 1952), a e ins umen al in bol-
s e ing lea ne me acogni ion and pe o mance moni o ing p ocesses.
Figu e 1 p o ides an o e iew o he co e condi ions o delibe a e p ac ice ou -
lined in his chap e . I does no ep esen an exhaus i e lis o all condi ions
men ioned in he Expe ise S udies li e a u e, bu a he ocuses on hose condi-
ions aken up in T ansla ion S udies o da e.
3 Challenges o he implemen a ion o delibe a e p ac ice
I is wo h no ing he e ha empi ical esea ch on he bene i s o delibe a e p ac-
ice on ansla ion pe o mance is s ill qui e scan . This dea h can a leas be
pa ly explained by a se ies o inhe en challenges associa ed wi h i s implemen-
a ion in o mal aining con ex s. As men ioned in he p e ious sec ion, se e al
p oponen s o delibe a e p ac ice sugges ha i s bene i s a e bes ealised in con-
ex s in ol ing lea ne s wo king one on one wi h indi idual aine s and in he
absence o a p e-se cu iculum (E icsson e al. 1993: 367). This ype o design
does no eadily align wi h he ashion in which ansla o s a e usually ained
o a numbe o di e en easons, s a ing wi h he inancial cons ain o need-
ing o hi e a pe sonal aine .
Ano he cons ain po en ially s anding in he way o mee ing se e al o he
co e condi ions o delibe a e p ac ice is he deg ee o lea ne he e ogenei y com-
monly ound in class oom-based aining con ex s. S uden s o ansla ion o en
ha e widely a ying le els o compe ence and expe ience, no o men ion di-
30
2 Gene a i e AI as a acili a o o delibe a e p ac ice in ansla o aining
immedia e,
in o ma i e
eedback
e o
co ec ion
oppo uni ies
" aine "
p esence
app op ia e
di icul y
le el
in insic
mo i a ion
conscious
pe o mance
moni o ing
Co e condi ions
o delibe a e
p ac ice
Figu e 1: Co e condi ions o delibe a e p ac ice
e se lea ning needs and in e es s. This can make i di icul o aine s o es-
ablish uly indi idualised lea ning objec i es and co esponding asks o mee
hem. The p oblem is exace ba ed when s uden en olmen s a e high. Feedback,
a guably he mos impo an dimension o delibe a e p ac ice, o en becomes
less de ailed and, ou o necessi y, much less immedia e. Pee eedback and sel -
eedback ac i i ies can help add ess his gap, bu immediacy, as a c i e ion o
p ac ice o be delibe a e, o en emains e y di icul o ob ain.
Beyond ime cons ain s, he delibe a e p ac ice condi ion o in o ma i e, im-
media e eedback migh no be me i he kind o eedback being p o ided is no
ans e able in he sense o applying o u u e ansla ion asks. O en imes, as-
sessmen ub ics a e used o ma k up e o s in acco dance wi h a ious ex ual
le els, such as g amma , wo d choice, and syn ax. I he eedback gi en pe ains
o pa e ns along hese lines ha a e applicable ac oss asks, i could be ega ded
as uly in o ma i e. I , on he o he hand, eedback simply consis s o ma king
up one-o e o s pe aining o i ems ha he ansla o may ne e encoun e
31
E ik Angelone
again in u u e ansla ions, such as an isola ed colloca ion e o , i is ela i ely
shallow and no pa icula ly in o ma i e.
Lea ne he e ogenei y in la ge coho se ings also p esen s challenges when
i comes o mee ing he delibe a e p ac ice condi ion o making su e asks a e be-
ing unde aken a an app op ia e le el o di icul y. T ansla ion p ac ice cou ses
a e o en in o med by aine in ui ion and p edic ion o app op ia e di icul y
le el, in u n based on such ace s as le el o s udy, in ake examina ion esul s,
a ious cogni i e p ocess me ics (Sun & Sh e e 2014), o p e ious s uden pe -
o mance and co esponding ‘ ich poin s’ (PACTE 2011). Howe e , s udies ha e
shown a dange o misalignmen be ween pe cei ed o p edic ed p oblems and
wha ac ually p o es o be p oblema ic when ansla ing (Angelone 2018). Wha
is assumed o be di icul is ac ually no necessa ily so, making a emp s a p e-
dic ing and se ing an app op ia e di icul y le el challenging, pa icula ly ac oss
a wide ange o s uden s wi h di e ging needs. Fu he mo e, he idea o ha ing
s uden s wo k ou side o hei com o zone, o a he pe iphe y o wha hey
can ealis ically accomplish, o he pu pose o ad ancing expe ise may con a-
dic app oaches o ansla o aining ha emb ace he p edic abili y o s aying
wi hin he lea ne ’s com o zone.
Acco ding o delibe a e p ac ice guidelines, he lea ne ’s in insic mo i a ion
needs o be cons an . Ha ing ansla ion s uden s wo k ou side o hei com o
zone in pedagogical con ex s, pa icula ly when g ades a e in ol ed, uns he
isk o hampe ing such mo i a ion. The ou comes o delibe a e p ac ice would
need o be gauged using me ics beyond o mal g ades, wi h a ocus on help-
ing lea ne s become mo e sel - e lec i e ansla o s. In insic mo i a ion would
come no so much om ge ing good g ades o doing well in a gi en cou se, bu
a he om seeing he bene i s o pu ing in ha d wo k and emb acing di icul y
in o de o become a be e ansla o .
4 Gene a i e AI as a acili a o o delibe a e p ac ice
Agains he backd op o he condi ions o delibe a e p ac ice p esen ed in Sec-
ion 2 and he cons ain s po en ially s anding in i s way, as ou lined in Sec ion 3,
he ques ion emains: how can we bes go abou acili a ing delibe a e p ac ice
in ansla o aining o pu poses o expe ise acquisi ion and ad ancemen ?
In pa icula , how can we es ablish he equisi e highly indi idualised, ‘coach’-
o ien ed app oach a he hea o delibe a e p ac ice? The pedagogical ea u es
o gene a i e AI, in p o iding eal- ime eedback and p omp -d i en in e ac ion
in a use -cen ed ashion, would seem o hold po en ial in his ega d, as will be
illus a ed h ough a se ies o conc e e scena ios in Sec ion 6.
32
2 Gene a i e AI as a acili a o o delibe a e p ac ice in ansla o aining
A he ime o w i ing, we a e s ill wi nessing he dawn o gene a i e AI as a
ehicle o op imising ansla ion, bo h in p o essional and pedagogical con ex s.
I s inc easingly ubiqui ous in eg a ion by LSP companies in p ojec wo k lows
has d awn a en ion o a need o a i icial in elligence (AI) li e acy (K üge 2023),
alongside MT li e acy and da a li e acy in a b oad sense. AI li e acy can be de-
ined as “a se o compe encies ha enables indi iduals o c i ically e alua e AI
echnologies, communica e and collabo a e e ec i ely wi h AI, and use AI as a
ool online, a home, and in he wo kplace” (Long & Mage ko 2020:2). In con-
junc ion wi h discussions o ansla o pe o mance, much o he discou se on
AI in elligence o da e has ocused on how gene a i e AI can be le e aged o a-
cili a e human-in- he-loop ansla ion, wi h an emphasis on he ansla ion p od-
uc . Ou side o se e al o he chap e s in his olume, ela i ely li le discussion
has been dedica ed o he po en ial bene i s o gene a i e AI as a con e sa ional
agen , ocusing less on gene a ing ansla ed con en , and mo e so on enabling
ansla o s o e lec on hei pe o mance and engage in ansla ion asks o a-
cili a e delibe a e p ac ice.
Ex ending on sociocul u al lea ning heo ies (Vygo sky 1965), some ha e come
o ega d a gene a i e AI ool like Cha GPT as a ‘mo e knowledgeable o he ’
(MKO), in essence aking on he ole o a pe sonalised aine ha can “lead
he lea ne om he zone o cu en de elopmen o he zone o p oximal de-
elopmen – he space whe e one canno qui e mas e a con en / ask o hei
own, bu hey can wi h he help o an expe ” (S ojano 2023: 2). As an MKO,
Cha GPT can add ess he a o emen ioned condi ions o delibe a e p ac ice, pa -
icula ly he p esence o a one-on-one pe sonalised aine . Th ough eal- ime
esponses o lea ne p omp s, Cha GPT ensu es he immediacy o eedback ha
is so di icul o ob ain in a la ge en olmen class oom-based ansla o aining
en i onmen . The lea ne ’s in insic mo i a ion is likely o be heigh ened when
aining is pe sonalised and sel -d i en, oo ed in immedia e, in o ma i e eed-
back, and in e ac i e, d i en by hei own p omp s in ela ion o aspec s o hei
own pe o mance.
5 Gene a i e AI as a sca old o sel -di ec ed lea ning
The u ilisa ion o gene a i e AI o pu poses o delibe a e p ac ice calls o he
lea ne o pa ake in sel -di ec ed lea ning (SDL), whe e “indi iduals ake he ini-
ia i e, wi h o wi hou he help o o he s, in diagnosing hei lea ning needs, se -
ing lea ning goals, iden i ying esou ces, choosing app op ia e lea ning s a e-
gies, and e alua ing hei lea ning ou comes” (Knowles 1975: 18). In his case,
33
E ik Angelone
gene a i e AI helps sca old lea ning in line wi h lea ne p omp s. T ansla ion
aine s well- e sed in he condi ions o delibe a e p ac ice can p o ide lea ne s
wi h aluable in o ma ion on he na u e o p omp s hey should en e . Howe e ,
sel -di ec ed lea ne s need o be “able, eady, and willing o p epa e, execu e
and comple e lea ning independen ly” (Jossbe ge e al. 2010: 419). The need o
lea ne independence does no make he aine supe luous, bu a he shi s he
ocus o assignmen s unde aken and how hey a e assessed.
Models o sel -di ec ed lea ning bea e y close esemblance o he delibe a e
p ac ice model. One such model ha is widely ci ed in he li e a u e consis s
o h ee closely in e ela ed dimensions: 1) sel -managemen , 2) sel -moni o ing,
and 3) mo i a ion (Ga ison 1997). Sel -managemen in ol es he lea ne es ab-
lishing conc e e lea ning goals and managing lea ning esou ces o achie e hese
goals. In o he wo ds, hey ake con ol, deciding on he asks in which hey
will engage. F om a delibe a e p ac ice pe spec i e, h ough s a egic p omp s,
he ansla o can le e age gene a i e AI o anno a e e o s in hei ansla ions.
Gene a ed anno a ions could hen se e as a amewo k o he ansla o o sel -
disco e he na u e o he e o s. Gene a i e AI could hen be p omp ed o p o-
ide simila ansla ion asks, wi h he goal o engaging he lea ne in delibe a e
p ac ice cen ed a ound a ce ain e o pa e n (such as a oiding alse cogna es,
e oneous li e al ansla ion, o p oblema ic ansla ionese a a syn ac ic le el).
Beyond e o de ec ion and mi iga ion, he ansla o can also use gene a i e AI
p omp ing o sel -manage he di icul y le el o he asks hey a e unde aking.
Sec ion 6 p o ides mo e conc e e scena ios and desc ip ions along hese lines.
Sel -moni o ing, he second componen o Ga ison’s model, pe ains o he
lea ne ’s me acogni i e p ocesses. As an impo an dimension o sel -di ec ed
lea ning, sel -moni o ing “ equi es lea ne s o ake esponsibili y o cons uc
meanings” (Ga ison 1997: 24). Th ough in e ac i e eedback, gene a i e AI can
shed aluable ligh o help lea ne s iden i y salien ea u es o he ansla ion ask
on which o ocus hei a en ion. Fo example, ansla o s can en e a p omp
asking gene a i e AI o anno a e sou ce con en ha could be an icipa ed o p e-
dic ed o p esen challenges in ansla ion. O e he pas decade, sc een eco ding
has ound a place in p ocess-o ien ed ansla o aining o pu poses o os e ing
sel -moni o ing and o enhance lea ne me acogni ion based on documen a ion
o ansla ion beha io s sugges ing p oblems, including pausing, in o ma ion e-
ie al, and e ision (Angelone 2019). A p esen , gene a i e AI ools like Cha -
GPT do no o e unc ionali y whe e ansla o s can upload sc een eco dings o
hei wo k o pu poses o ecei ing analy ic eedback a a g anula le el. Gi en
ecen ad ancemen s in his echnology howe e , such as APIs ha can p o ide
34
2 Gene a i e AI as a acili a o o delibe a e p ac ice in ansla o aining
au oma ed ideo summa isa ion, i is qui e likely ha u ilisa ion o gene a i e
AI ools o such pu poses is no oo a away.
The hi d componen o Ga ison’s sel -di ec ed lea ning model, mo i a-
ion, di ec ly aligns wi h mo i a ion as a co e condi ion o delibe a e p ac ice.
Whe eas delibe a e p ac ice ega ds mo i a ion as he lea ne emb acing chal-
lenge h ough an inhe en desi e o become be e , Ga ison d aws a en ion o
he impo ance o mo i a ion o pu poses o s aying on ask. The con e sa ional
in e ace o gene a i e AI ools such as Cha GPT equi es ac i e pa icipa ion
on behal o he lea ne , and, he eby, a heigh ened need o s ay on ask. How-
e e , wi hou he physical p esence o ac ual aine s o pee s, s aying on ask,
and mo i a ion in gene al, is no a gi en. Cha GPT is no inclined o openly
p aise he pe o mance o he lea ne o pique in e es and mo i a ion. Indeed,
a p esen , he Cha GPT in e ace i sel is qui e basic, lacking any s uc u al o
discou se elemen s ha migh lend hemsel es well o inhe en ly acili a ing
lea ne mo i a ion. I will be in e es ing o see i his changes o e ime, pe haps
in line wi h empi ical use expe ience s udies.
6 Applica ion scena ios
This chap e will now p o ide a se ies o conc e e scena ios o illus a e how
gene a i e AI can be used o acili a e each o he co e dimensions o delibe -
a e p ac ice pu o wa d in Sec ion 2 and o help ansla o s ad ance along an
expe ise ajec o y. The examples will be based on in e ac ion wi h Cha GPT
based on GPT-4o,1gi en he ela i e ubiqui y and popula i y o his pa icula
gene a i e AI ool a he ime o w i ing. This ocus on Cha GPT o illus a-
i e pu poses does no disc edi he g owing ange o o he gene a i e AI ools
a ailable o use in a simila ashion. The ansla ion scena io being used o
pu poses o con ex ualisa ion is he Ge man-English ansla ion o web con en
om a Ge man p i a e heal h insu ance company,2 ansla ed o an in e na-
ional, English-speaking audience o in o ma i e pu poses. The English ansla-
ion was gene a ed using DeepL.3
Figu e 2 p o ides side-by-side alignmen o he sou ce and a ge con en .
As a poin o depa u e, and o unde sco e he impo ance o s a egic p omp -
ing when i comes o u ilising GenAI o acili a ing delibe a e p ac ice, (2) de-
1h ps://openai.com/index/gp -4/
2h ps://www.huk.de/gesundhei - o so ge- e moegen/k anken e siche ung/
k anken oll e siche ung.h ml#huk
3h ps://www.deepl.com/en/ ansla o
35
E ik Angelone
(11) “P o ide me wi h a Ge man sou ce ex o ansla e in his same domain
ha is a li le mo e di icul han my p e ious ansla ion”.
(12) Gene a ion o a sligh ly mo e di icul ansla ion ask
Ih e p i a e K anken e siche ung de HUK-COBURG
Wa um sich eine p i a e K anken e siche ung lohn
• Um ang eiche Schu z – Unse e Ta i e bie en um assende
Leis ungen, die wei übe das hinausgehen, was gese zliche
K ankenkassen abdecken.
• Schnelle Te min e gabe – Be o zug e Behandlung bei Fachä z en
und kü ze e Wa ezei en.
• Hohe Flexibili ä – Wählen Sie aus e schiedenen
Selbs be eiligungss u en und Ta i op ionen, die sich Ih en
indi iduellen Bedü nissen anpassen.
• P ä en ionsangebo e – Zugang zu exklusi en
Gesundhei sp og ammen und Vo so geun e suchungen.
• In e na ionale Schu z – Wel wei e Absiche ung ü Reisen und
Au en hal e im Ausland.
• T anspa en e Bei äge – Keine e s eck en Kos en, kla e und
nach ollziehba e P ämienbe echnung.
Es ablishing an app op ia e di icul y le el o en p o es o be a daun ing ask
o aine s in o mal ansla ion class oom se ings. I is in e es ing o see how
Cha GPT goes abou desc ibing wha exac ly i is ha makes his ask mo e
di icul when p omp ed o do so, as e lec ed in (13).
(13) Cha GPT’s desc ip ion o wha makes he p oposed ask mo e di icul
This ansla ion ask is mo e di icul because i includes:
•Complex Vocabula y and Ph asing: The Ge man ex uses mo e
speci ic and echnical e ms ela ed o insu ance, such as
"um ang eiche Schu z," "Selbs be eiligungss u en," and
"P ämienbe echnung." T ansla ing hese e ms accu a ely
equi es a good unde s anding o insu ance e minology in
bo h languages.
•Longe Sen ences and Mo e De ailed Con en : The sen ences
a e longe and con ain mo e de ailed in o ma ion, which
equi es ca e ul a en ion o main ain cla i y and
cohe ence in English.
42
2 Gene a i e AI as a acili a o o delibe a e p ac ice in ansla o aining
•Va ied Bene i s and Fea u es: The ex lis s mul iple
bene i s and ea u es o he insu ance, each wi h i s own
nuances. Ensu ing ha each poin is clea ly and accu a ely
ansla ed wi hou losing meaning o con ex adds o he
complexi y.
•Idioma ic Exp essions: Ce ain ph ases in Ge man, such as
"wei übe das hinausgehen" and "be o zug e Behandlung,"
migh no ha e di ec equi alen s in English, equi ing
he ansla o o ind idioma ic exp essions ha con ey
he same meaning.
A a quick glance, much o wha is deemed o be “mo e di icul ” is somewha
gene ic (mo e complex ield-speci ic e minology, a longe ex , po en ial lexical
gaps, e c.), as opposed o pe sonalised o he indi idual ansla o based on p e-
ious pe o mance. This condi ion o delibe a e p ac ice may e y well p o e o
be he mos challenging o no only human aine s, bu also Cha GPT o mee .
Pe haps Cha GPT’s e icacy in doing so hinges on he ansla o en e ing mul-
iple ansla ions as pa o he p omp . These ansla ions could hen se e as
a co pus o ‘big da a’ o aining he sys em o base wha is being p oposed as
mo e di icul on conc e e pa e ns (e o s, ex a ibu es, e c.).
7.4 In insic mo i a ion
The ex en o which lea ne s a e in insically mo i a ed o pa ake in he scena -
ios ou lined in his chap e is in e wined wi h he ex en o which hey alue and
us Cha GPT as a ‘ aine ’ and he ex en o which hey emb ace sel -di ec ed
lea ning. Bo h can ake some ime and necessi a e adjus men , pa icula ly in
he case o ansla ion s uden s who migh be used o in e ac ing wi h pee s
and ins uc o s. As p e iously men ioned, in insic mo i a ion in he con ex o
sel -di ec ed, gene a i e AI-based delibe a e p ac ice will likely s em i s and
o emos om he le el o pe sonalisa ion and one-on-one in e ac ion he ech-
nology enables. Lea ne s a e e y much in he d i e ’s sea . I is hei p omp ing
and con en inpu ha se es as a ca alys o he en i e delibe a e p ac ice p o-
cess. Ne e heless, in insic mo i a ion could quickly al e in ins ances whe e
gene a i e AI makes e o s o p o ides unhelp ul eedback. To main ain mo i a-
ion, ansla o s would need o look beyond such isola ed ins ances and see he
bigge pic u e o how he echnology can os e hei expe ise acquisi ion on he
whole. Depending on lea ning s yles and p e e ences, some lea ne s may e en
ind mo i a ion a he idea o being able o openly ques ion and challenge he
43
E ik Angelone
non-human ‘ aine ’ in such con ex s, and would pe haps be less inclined o do
so in con ex s in ol ing ace- o- ace ins uc ion in ol ing human ainees and
aine s.
7.5 Conscious pe o mance moni o ing
Success ul delibe a e p ac ice calls o he lea ne o be consciously engaged in
e lec ion on hei pe o mance. In he absence o an ac ual human aine and
ace- o- ace lea ning, i is basically up o he lea ne o make su e hey a e on
ask and ac i ely engaged in such moni o ing. This impo an condi ion o delib-
e a e p ac ice may p o e o be di icul o Cha GPT o acili a e in a ashion ha
is inhe en ly ad an ageous in ela ion o wha one migh ind in a ace- o- ace
lea ning en i onmen . When p omp ed o desc ibe ways in which i can acili-
a e lea ne s’ conscious pe o mance moni o ing when u ilised as a sole ‘ aine ’,
Cha GPT p oposed such s a egies as p og amming he ool o pe iodically ask
use s o eedback on hei pe o mance, ye also poin s ou ha i could no ack
de ailed usage analy ics o goal p og ession. Ins ead, i can guide use s on how
o ack hei usage manually. In o he wo ds, he onus is uly on he lea ne
when i comes o mee ing his condi ion o delibe a e p ac ice.
8 Conclusion
Whe eas many o he cu en discussions su ounding gene a i e AI in ol e i s
po en ial place in ansla ion p ojec wo k lows and in he gene a ion o ans-
la ed con en , his chap e ad oca es o le e aging i s pedagogical ea u es as
an MKO and con e sa ional agen in acili a ing delibe a e p ac ice and expe -
ise acquisi ion o human ansla o s. I s p ima y ad an ages in his capaci y
lie in i s abili y o gene a e immedia e, in o ma i e eedback in a highly pe son-
alised ashion. As a as he na u e o eedback is conce ned, a ool like Cha GPT
is equally e icacious a anno a ion o pu poses o e o co ec ion and desc ip-
ion o pu poses o acili a ing nuanced unde s anding and e o mi iga ion. I
can eadily p o ide ansla o s wi h delibe a e p ac ice in ela ion o a gi en ex
a ibu e o e o ype, wi h an eye owa ds ans e abili y ac oss asks. Pe son-
alised eedback, one-on-one aining, and ongoing in e ac ion h ough lea ne
p omp en y and ool esponse can os e in insic mo i a ion. The co e delib-
e a e p ac ice condi ions o making su e asks a e unde aken a an app op ia e
di icul y le el and ha lea ne s engage in conscious pe o mance moni o ing,
a leas a he ime o w i ing, a e s ill somewha o a challenge o Cha GPT, ye
bo h can be add essed h ough s a egic p omp ing.
44
2 Gene a i e AI as a acili a o o delibe a e p ac ice in ansla o aining
T uly expe imen al s udies on he bene i s o sel -di ec ed, delibe a e p ac ice
in ela ion o mo e adi ional lea ning app oaches a e s ill limi ed in scope, o
he main eason ha he co e condi ions o delibe a e p ac ice ha e been qui e
ha d o add ess o da e (Mille e al. 2020: 8). This is no some hing speci ic o
ansla o aining, bu aining in gene al, in all si ua ions whe e lea ne he e o-
genei y, ime cons ain s, and he absence o one-on-one aining oppo uni ies
eme ge as oadblocks. Cha GPT o e s un apped po en ial o be a game change
in his ega d and will hope ully d aw enewed in e es in bo h pedagogical and
empi ical esea ch on delibe a e p ac ice and i s place in ansla ion expe ise
acquisi ion.
Re e ences
Angelone, E ik. 2018. Reconcep ualizing p oblems in ansla ion using iangu-
la ed p ocess and p oduc da a. In Rii a Jääskeläinen & Isabel Lac uz (eds.),
Inno a ion and expansion in ansla ion p ocess esea ch, 17–36. Ams e dam,
Ne he lands: John Benjamins.
Angelone, E ik. 2019. P ocess-o ien ed assessmen o p oblems and e o s in
ansla ion: Expanding ho izons h ough sc een eco ding. In Elsa Hue as-
Ba os, Sonia Vandepi e & Emilia Iglesias-Fe nández (eds.), Quali y assu ance
and assessmen p ac ices in ansla ion and in e p e ing, 179–198. He shey, USA:
IGI Global.
Angelone, E ik. 2023. Wea ing adap i e expe ise in o ansla o aining. In
Ga y Massey, Elsa Hue as-Ba os & Da id Ka an (eds.), The human ansla o
in he 2020s, 60–73. London, UK: Rou ledge.
D ey us, Hube & S ua D ey us. 1986. Mind o e machine: The powe o human
in ui ion and expe ise in he e a o he compu e . Ox o d, UK: Basil Blackwell.
EMT. 2022. Eu opean mas e ’s in ansla ion compe ence amewo k 2022. Tech.
ep. B ussels: Eu opean Commission. 1–12. h ps://commission.eu opa.eu/
sys em/ iles/2022-11/em _compe ence_ wk_2022_en.pd .
E icsson, K. Ande s. 2006. The in luence o expe ience and delibe a e p ac ice
on he de elopmen o supe io expe pe o mance. In The Camb idge hand-
book o expe ise and expe pe o mance, 683–703. Camb idge, UK: Camb idge
Uni e si y P ess. DOI: 10.1017/CBO9780511816796.038.
E icsson, K. Ande s & Neil Cha ness. 1994. Expe pe o mance: I s s uc u e and
acquisi ion. Ame ican Psychologis 49. 725–747.
E icsson, K. Ande s, Ral T. K ampe & Clemens Tesch-Röme . 1993. The ole
o delibe a e p ac ice in he acquisi ion o expe pe o mance. Psychological
Re iew 100(3). 363–406. DOI: 10.1037/0033-295X.100.3.363.
45
E ik Angelone
E icsson, K. Ande s & And eas C. Lehmann. 1996. Expe and excep ional pe o -
mance: E idence o maximal adap a ion o ask cons ain s. Annual Re iew o
Psychology 47. 273–305. DOI: 10.1146/annu e .psych.47.1.273.
Ga ison, D. R. 1997. Sel -di ec ed lea ning: Towa d a comp ehensi e model.
Adul Educa ion Qua e ly 48(1). 18–33. DOI: 10.1177/074171369704800103.
Gladwell, Malcolm. 2011. Ou lie s: The s o y o success. New Yo k, USA: Back Bay
Books.
Ha ano, Giyoo & Kayako Inagaki. 1986. Two cou ses o expe ise. In Ha old
S e enson, Hi oshi Azuma & Kenji Haku a (eds.), Child de elopmen and ed-
uca ion in Japan, 262–272. New Yo k, USA: W. Y. F eeman & Co.
Ho n, John & Hi omi Masunaga. 2006. A me ging heo y o expe ise and in elli-
gence. In The Camb idge handbook o expe ise and expe pe o mance, 587–611.
Camb idge, UK: Camb idge Uni e si y P ess. DOI: 10.1017/CBO9780511816796.
034.
Jossbe ge , Helen, Saskia B and-G uwel, Henny Boshuizen & Ma gje an de Wiel.
2010. The challenge o sel -di ec ed and sel - egula ed lea ning in oca ional
educa ion: A heo e ical analysis and syn hesis o equi emen s. Jou nal o
Voca ional Educa ion & T aining 62(4). 415–440. DOI: 10.1080/13636820.2010.
523479.
Knowles, Malcolm. 1975. Sel -di ec ed lea ning: A guide o lea ne s and eache s.
Chicago, USA: Folle .
K üge , Ralph. 2023. A i icial in elligence li e acy o he language indus y
– wi h pa icula emphasis on ecen la ge language models such as GPT-4.
Lebende Sp achen 68(2). 283–330. DOI: 10.1515/les-2023-0024.
Long, Du i & B ian Mage ko. 2020. Wha is AI li e acy? Compe encies and design
conside a ions. In P oceedings o he 2020 CHI Con e ence on Human Fac o s in
Compu ing Sys ems (CHI ’20), 1–16. Honolulu, USA: Associa ion o Compu -
ing Machine y. DOI: 10.1145/3313831.3376727. h ps://doi.o g/10.1145/3313831.
3376727.
Mille , Sco D., Da yl Chow, B uce E. Wampold, Ma k A. Hubble, A. C. Del
Re, Cyn hia Maeschalck & Susanne Ba gmann. 2020. To be o no o be (an
expe )? Re isi ing he ole o delibe a e p ac ice in imp o ing pe o mance.
High Abili y S udies 31(1). 5–15. DOI: 10.1080/13598139.2018.1519410. h ps:
//doi.o g/10.1080/13598139.2018.1519410.
PACTE. 2011. Resul s o he alida ion o he PACTE ansla ion compe ence
model: T ansla ion p oblems and ansla ion compe ence. In Cecilia Al s-
ad, Adelina Hild & Elisabe Tiselius (eds.), Me hods and s a egies o p ocess
esea ch: In eg a i e app oaches o ansla ion s udies, 317–343. Ams e dam,
Ne he lands: John Benjamins.
46
2 Gene a i e AI as a acili a o o delibe a e p ac ice in ansla o aining
Piage , Jean. 1952. The o igins o in elligence in child en. New Yo k, USA: In e na-
ional Uni e si ies P ess.
Sh e e, G ego y. 2006. The delibe a e p ac ice: T ansla ion and expe ise. Jou nal
o T ansla ion S udies 9(1). 27–42.
Sh e e, G ego y. 2018. Le els o explana ion and ansla ion expe ise. He mes 57.
97–108.
Sh e e, G ego y. 2019. P o essional ansla o de elopmen om an expe ise
pe spec i e. In E ik Angelone, Mau een Eh ensbe ge -Dow & Ga y Massey
(eds.), The Bloomsbu y companion o language indus y s udies. London, UK:
Bloomsbu y Academic.
S ojano , Ana. 2023. Lea ning wi h Cha GPT 3.5 as a mo e knowledgeable o he :
An au oe hnog aphic s udy. In e na ional Jou nal o Educa ional Technology in
Highe Educa ion 20(1). 35. DOI: 10.1186/s41239-023-00404-7.
Sun, Sanjun & G ego y Sh e e. 2014. Measu ing ansla ion di icul y: An empi -
ical s udy. Ta ge 26(1). 98–127.
Vygo sky, Le . 1965. Though and language. Camb idge, USA: MIT P ess.
47
